Podcasts about convolutional neural networks

vision cnn llc copy improved modeling incorporating blur tong robustness biorxiv convolutional neural networks

Play Episode Listen Later Jul 31, 2023

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2023.07.29.551089v1?rss=1 Authors: Jang, H., Tong, F. Abstract: Whenever a visual scene is cast onto the retina, much of it will appear degraded due to poor resolution in the periphery; moreover, optical defocus can cause blur in central vision. However, the pervasiveness of blurry or degraded input is typically overlooked in the training of convolutional neural networks (CNNs). We hypothesized that the absence of blurry training inputs may cause CNNs to rely excessively on high spatial frequency information for object recognition, thereby causing systematic deviations from biological vision. We evaluated this hypothesis by comparing standard CNNs with CNNs trained on a combination of clear and blurry images. We show that blur-trained CNNs outperform standard CNNs at predicting neural responses to objects across a variety of viewing conditions. Moreover, blur-trained CNNs acquire increased sensitivity to shape information and greater robustness to multiple forms of visual noise, leading to improved correspondence with human perception. Our results provide novel neurocomputational evidence that blurry visual experiences are very important for conferring robustness to biological visual systems. Copy rights belong to original authors. Visit the link for more info Podcast created by Paper Player, LLC

We Are Designed to React Without Thinking: David Eravaci

Roll Call with Chappy

Play Episode Listen Later May 8, 2023 37:09

David's underprivileged upbringing led him down a path of rebellion and hardship. Despite this, David found a way to tap into his true self through his self-actualization to overcome his past and find peace in his presence. David firmly believes that we as people are designed to react without thinking. This can lead people into a world of anxiety, chaos, and reactionary minds. Therefore, David believes if you can control how you react to people, places, and things then you can control your life. Presently David works in Artificial Intelligence and Machine Learning with a focus on financial technology, natural language processing/language models, and Convolutional Neural Networks.In this episode, learn how David went from a life of destruction and pain to working in one of today's most prominent and controversial fields of AI. Find out how he released himself from a life that was no longer serving him, to using his incredible and intelligent brain to change his circumstances for the better.Connect with Peter Meyerhoff: · Website · Instagram· TikTok· Facebook· Email: petermeyerhoff.info@gmail.comConnect with David Eravaci:· Instagram

ai thinking artificial intelligence designed react machine learning convolutional neural networks peter meyerhoff

What Affects Array Language Performance?

The Array Cast

Play Episode Listen Later Mar 18, 2023 85:17

Array Cast - March 17, 2023 Show NotesThanks to Bob Therriault, Richard Park, Conor Hoekstra and Adám Brudzewsky for gathering these links:[01] 00:01:55 APL problem solving competition https://contest.dyalog.com/ Kattis online competition https://open.kattis.com/ APL Seeds '23 https://www.dyalog.com/apl-seeds-user-meetings/aplseeds23.htm Linux Format Magazine https://linuxformat.com/linux-format-300.html The APL Show - Reaction to "Change the Way You Think" https://apl.show/2023/03/09/Reaction-to-Change-the-way-you-write-Change-the-way-you-think-part-1.html The APL Campfire - Norman Thomson https://www.youtube.com/watch?v=jPujK-GvHGQ&list=PLYKQVqyrAEj91hZHbJiWOENHZP4JT8VFv[02] 00:06:16 Ed Gottsman's Wiki Gui https://www.youtube.com/watch?v=j17E_KUgKxk[03] 00:07:09 Why I Love BQN So Much https://www.youtube.com/watch?v=mRT-yK2RTdg J software https://www.jsoftware.com/#/ Dyalog APL https://www.dyalog.com/[04] 00:08:12 Adám's APL Quest https://www.youtube.com/@abrudz/playlists[05] 00:09:50 q download https://kx.com/kdb-personal-edition-download/[06] 00:13:10 Shakti https://shakti.com/[07] 00:14:10 Emery Berger "Performance Really Matters" https://www.youtube.com/watch?v=7g1Acy5eGbE[08] 00:17:14 Three consecutive odds ADSP 'scanductions' episode https://adspthepodcast.com/2023/03/03/Episode-119.html[09] 00:19:40 Rich Park's "A Programming Language for Thinking About Algorithms" https://www.dyalog.com/uploads/files/presentations/ACCU20210520.pdf[10] 00:21:00 Windows function in BQN https://mlochbaum.github.io/BQN/doc/windows.html[11] 00:27:22 Fold in J https://code.jsoftware.com/wiki/Vocabulary/fcap Scan https://aplwiki.com/wiki/Scan Reduce https://aplwiki.com/wiki/Reduce[12] 00:29:15 Apex Compiler https://gitlab.com/bernecky/apex Co-dfns Compiler https://dl.acm.org/doi/10.1145/2627373.2627384[13] 00:32:50 Arthur Whitney https://en.wikipedia.org/wiki/Arthur_Whitney_(computer_scientist)[14] 00:37:03 Convolutional Neural Networks https://dl.acm.org/doi/pdf/10.1145/3315454.3329960[15] 00:39:05 Tensorflow https://en.wikipedia.org/wiki/Tensorflow PyTorch https://en.wikipedia.org/wiki/Pytorch MLIR https://mlir.llvm.org/[16] 00:44:20 Paul Graham "Beating the Averages" http://www.paulgraham.com/avg.html Bob Bernecky "Good Algorithms Win Over Tin" https://code.jsoftware.com/wiki/Essays/GoodAlgorithmsWinOverTin cudnn: https://developer.nvidia.com/cudnn C++/Python Meme https://www.reddit.com/r/ProgrammerHumor/comments/m3pf9h/there_is_only_one_king/[17] 00:49:00 Futhark Episode of ArrayCast https://www.arraycast.com/episodes/episode37-futhark Single Assignment C https://www.sac-home.org/index Dex https://github.com/google-research/dex-lang#dex-[18] 01:06:40 BQN Compiler https://mlochbaum.github.io/BQN/implementation/bootbench.html[19] 01:13:19 BQN Performance https://mlochbaum.github.io/BQN/implementation/perf.html Bench Array https://mlochbaum.github.io/bencharray/pages/summary.html[20] 01:16:12 Big Endian https://en.wikipedia.org/wiki/Endianness[21] 01:21:45 Performance Timing BQN _timed https://mlochbaum.github.io/BQN/spec/system.html#time J 6!:2 https://code.jsoftware.com/wiki/Vocabulary/Foreigns#m6 APL cmpx http://dfns.dyalog.com/n_cmpx.htm q ts:n https://code.kx.com/q/basics/syscmds/#ts-time-and-space[22] 01:23:15 ngn/k https://codeberg.org/ngn/k[23] 01:23:52 Contact AT ArrayCast DOT Com

Representing Multiple Visual Objects in the Human Brain and Convolutional Neural Networks

cnn llc visual copy objects representing human brain fmri chun xu jeong biorxiv convolutional neural networks

Play Episode Listen Later Mar 1, 2023

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2023.02.28.530472v1?rss=1 Authors: Mocz, V., Jeong, S., Chun, M., Xu, Y. Abstract: Objects in the real world often appear with other objects. To recover the identity of an object whether or not other objects are encoded concurrently, in primate object-processing regions, neural responses to an object pair have been shown to be well approximated by the average responses to each constituent object shown alone, indicating the whole is equal to the average of its parts. This is present at the single unit level in the slope of response amplitudes of macaque IT neurons to paired and single objects, and at the population level in response patterns of fMRI voxels in human ventral object processing regions (e.g., LO). Here we show that averaging exists in both single fMRI voxels and voxel population responses in human LO, with better averaging in single voxels leading to better averaging in fMRI response patterns, demonstrating a close correspondence of averaging at the fMRI unit and population levels. To understand if a similar averaging mechanism exists in convolutional neural networks (CNNs) pretrained for object classification, we examined five CNNs with varying architecture, depth and the presence/absence of recurrent processing. We observed averaging at the CNN unit level but rarely at the population level, with CNN unit response distribution in most cases did not resemble human LO or macaque IT responses. The whole is thus not equal to the average of its parts in CNNs, potentially rendering the individual objects in a pair less accessible in CNNs during visual processing than they are in the human brain. Copy rights belong to original authors. Visit the link for more info Podcast created by Paper Player, LLC

Fully automated detection of dendritic spines in 3D live cell imaging data using deep convolutional neural networks

deep data 3d llc copy detection imaging spines bissen fully automated biorxiv eppler dendritic rumpel convolutional neural networks

Play Episode Listen Later Jan 8, 2023

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2023.01.08.522220v1?rss=1 Authors: Vogel, F. W., Alipek, S., Eppler, J.-B., Triesch, J., Bissen, D., Acker-Palmer, A., Rumpel, S., Kaschube, M. Abstract: Dendritic spines are considered a morphological proxy for excitatory synapses, rendering them a target of many different lines of research. Over recent years, it has become possible to image simultaneously large numbers of dendritic spines in 3D volumes of neural tissue. In contrast, currently no automated method for spine detection exists that comes close to the detection performance reached by human experts. However, exploiting such datasets requires new tools for the fully automated detection and analysis of large numbers of spines. Here, we developed an efficient analysis pipeline to detect large numbers of dendritic spines in volumetric fluorescence imaging data. The core of our pipeline is a deep convolutional neural network, which was pretrained on a general-purpose image library, and then optimized on the spine detection task. This transfer learning approach is data efficient while achieving a high detection precision. To train and validate the model we generated a labelled dataset using five human expert annotators to account for the variability in human spine detection. The pipeline enables fully automated dendritic spine detection and reaches a near human-level detection performance. Our method for spine detection is fast, accurate and robust, and thus well suited for large-scale datasets with thousands of spines. The code is easily applicable to new datasets, achieving high detection performance, even without any retraining or adjustment of model parameters. Copy rights belong to original authors. Visit the link for more info Podcast created by Paper Player, LLC

Choosing an Array Language / The Games We Play

The Array Cast

Play Episode Listen Later Jan 7, 2023 82:59

Array Cast - January 6, 2023 Show NotesThanks to Bob Therriault, Adám Brudzewsky, and Marshall Lochbaum for gathering these links:[01] 00:01:13 Twitter Poll for APL Cast https://twitter.com/a_brudz/status/1607653845445873664[02] 00:04:30 Revamped BQNcrate https://mlochbaum.github.io/bqncrate/[03] 00:06:44 APLcart https://aplcart.info[04] 00:07:43 Inclusive Range in Q https://www.5jt.com/the-rest-is-silence p: Prime in J https://code.jsoftware.com/wiki/Vocabulary/pco Prime in Dyalog APL https://dfns.dyalog.com/n_pco.htm[05] 00:09:42 Consecutive values https://mlochbaum.github.io/bqncrate/?q=consecutive%20values[06] 00:11:46 APL Tacit help https://tacit.help BQN https://saltysylvi.github.io/bqn-tacit-helper/ J tte tacit to explicit https://code.jsoftware.com/wiki/Addons/debug/tte 13 : explicit to tacit https://code.jsoftware.com/wiki/Vocabulary/com J Phrases https://code.jsoftware.com/wiki/Phrases[07] 00:19:39 Fun Q https://fun-q.net/ APL Farm Discord/Matrix https://apl.wiki/APL_Farm[08] 00:22:00 Nick Psaris Episode on ArrayCast https://www.arraycast.com/episodes/episode42-nick-psaris-q[09] 00:24:20 Extended Precision and Rational Types in J https://www.jsoftware.com/help/jforc/elementary_mathematics_in_j.htm#_Toc191734516 BQN systemMath.fact https://github.com/mlochbaum/BQN/blob/master/spec/system.md#math NARS 2000 https://aplwiki.com/wiki/NARS2000[10] 00:26:55 Dyalog Licence https://www.dyalog.com/prices-and-licences.htm CBQN GPL-3 Licence https://github.com/dzaima/CBQN#license J GPL-3 Licence https://github.com/jsoftware/jsource/blob/master/license.txt q Licence https://kx.com/developers/download-licenses/[11] 00:29:05 April Programming Language https://aplwiki.com/wiki/April[12] 00:31:20 Sort in BQN https://github.com/mlochbaum/BQN/blob/master/doc/order.md#sort Without in APL https://aplwiki.com/wiki/Without Less in J https://code.jsoftware.com/wiki/Vocabulary/minusdot#dyadic[13] 00:34:30 Jelly programming language https://apl.wiki/Jelly https://github.com/DennisMitchell/jellylanguage[14] 00:35:08 Rust programming language https://www.rust-lang.org/[15] 00:36:40 Lesser of >. in J https://code.jsoftware.com/wiki/Vocabulary/ltdot#dyadic[16] 00:38:20 Code Golf https://apl.wiki/Code_golf Parse float functionhttps://mlochbaum.github.io/BQN/spec/system.html#input-and-output[17] 00:40:44 APL ⎕D https://help.dyalog.com/latest/#Language/System%20Functions/d.htm APL ⎕C https://help.dyalog.com/latest/#Language/System%20Functions/c.htm APL ⎕A https://help.dyalog.com/latest/#Language/System%20Functions/a.htm Advent of Code https://en.wikipedia.org/wiki/Advent_of_Code[18] 00:43:16 APLx https://aplwiki.com/wiki/APLX APL PLUS https://aplwiki.com/wiki/APL*PLUS[19] 00:46:23 Dyalog ⎕DT https://help.dyalog.com/latest/#Language/System%20Functions/dt.htm[20] 00:52:46 Jelly Tutorial https://github.com/DennisMitchell/jellylanguage/wiki/Tutorial[21] 00:57:10 Plus Scan in BQN https://github.com/mlochbaum/BQN/blob/master/doc/scan.md APL +.× https://help.dyalog.com/latest/#Language/Primitive%20Operators/Inner%20Product.htm J +/ . * https://www.jsoftware.com/help/jforc/applied_mathematics_in_j.htm#_Toc191734505[22] 01:00:30 q advent of code solutions http://github.com/qbists/studyq/[23] 01:01:30 SQL https://en.wikipedia.org/wiki/SQL q for Mortals https://code.kx.com/q4m3/[24] 01:04:21 BQN Advent of Code list https://mlochbaum.github.io/BQN/community/aoc.html[25] 01:08:42 Adám's link http://www.jsfuck.com/ https://en.wikipedia.org/wiki/JSFuck[26] 01:10:02 q links for Advent of Code https://github.com/qbists/studyq/tree/main/aoc/2022 J forums Advent of Code https://www.jsoftware.com/cgi-bin/forumsearch.cgi?all=&exa=advent+of+code&one=&exc=&add=&sub=&fid=&tim=0&rng=0&dbgn=1&mbgn=1&ybgn=2005&dend=31&mend=12¥d=2022 J wiki Advent of Code https://code.jsoftware.com/wiki/Essays/Advent_Of_Code APL wiki Advent of Code https://apl.wiki/aoc K Wiki Advent of Code: https://k.miraheze.org/wiki/Advent_of_Code[27] 01:12:40 Convolutional Neural Networks in APL https://dl.acm.org/doi/pdf/10.1145/3315454.3329960 Neural Networks https://aplwiki.com/wiki/Neural_networks[28] 01:15:00 Dr. Raymond Polivka's new APL book: http://aplclass.com/book/ APL Stefan Kruger Learning APL https://aplwiki.com/wiki/Books#Learning_APL J J for C Programmers https://www.jsoftware.com/help/jforc/contents.htm J Playground Example|Neural Networks https://jsoftware.github.io/j-playground/bin/html2/# BQN Tutorials https://mlochbaum.github.io/BQN/tutorial/index.html[29] 01:17:38 APL Wiki Learning Resources https://aplwiki.com/wiki/Learning_resources k Wiki Learning Resources https://k.miraheze.org/wiki/Learning_Resources J Wiki Learning Resources https://code.jsoftware.com/wiki/Guides/GettingStarted[30] 01:19:21 Contact AT ArrayCast DOT com

Emergence of brain-like mirror-symmetric viewpoint tuning in convolutional neural networks

learning brain 3d llc mirror copy tuning emergence viewpoint golan zarco biorxiv symmetric convolutional neural networks

Play Episode Listen Later Jan 5, 2023

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2023.01.05.522909v1?rss=1 Authors: Farzmahdi, A., Zarco, W., Freiwald, W., Kriegeskorte, N., Golan, T. Abstract: Primates can recognize objects despite 3D geometric variations such as in-depth rotations. The computational mechanisms that give rise to such invariances are yet to be fully understood. A curious case of partial invariance occurs in the macaque face-patch AL and in fully connected layers of deep convolutional networks in which neurons respond similarly to mirror-symmetric views (e.g., left and right profiles). Why does this tuning develop? Here, we propose a simple learning-driven explanation for mirror-symmetric viewpoint tuning. We show that mirror-symmetric viewpoint tuning for faces emerges in the fully connected layers of convolutional deep neural networks trained on object recognition tasks, even when the training dataset does not include faces. First, using 3D objects rendered from multiple views as test stimuli, we demonstrate that mirror-symmetric viewpoint tuning in convolutional neural network models is not unique to faces: it emerges for multiple object categories with bilateral symmetry. Second, we show why this invariance emerges in the models. Learning to discriminate among bilaterally symmetric object categories induces reflection-equivariant intermediate representations. AL-like mirror-symmetric tuning is achieved when such equivariant responses are spatially pooled by downstream units with sufficiently large receptive fields. These results explain how mirror-symmetric viewpoint tuning can emerge in neural networks, providing a theory of how they might emerge in the primate brain. Our theory predicts that mirror-symmetric viewpoint tuning can emerge as a consequence of exposure to bilaterally symmetric objects beyond the category of faces, and that it can generalize beyond previously experienced object categories. Copy rights belong to original authors. Visit the link for more info Podcast created by Paper Player, LLC

The performance of the MAGIC telescopes using deep convolutional neural networks with CTLearn

Play Episode Listen Later Nov 30, 2022 0:49

The performance of the MAGIC telescopes using deep convolutional neural networks with CTLearn by T. Miener et al. on Wednesday 30 November The Major Atmospheric Gamma Imaging Cherenkov (MAGIC) telescope system is located on the Canary Island of La Palma and inspects the very high-energy (VHE, few tens of GeV and above) gamma-ray sky. MAGIC consists of two imaging atmospheric Cherenkov telescopes (IACTs), which capture images of the air showers originating from the absorption of gamma rays and cosmic rays by the atmosphere, through the detection of Cherenkov photons emitted in the shower. The sensitivity of IACTs to gamma-ray sources is mainly determined by the ability to reconstruct the properties (type, energy, and arrival direction) of the primary particle generating the air shower. The state-of-the-art IACT pipeline for shower reconstruction is based on the parameterization of the shower images by extracting geometric and stereoscopic features and machine learning algorithms like random forest or boosted decision trees. In this contribution, we explore deep convolutional neural networks applied directly to the pixelized images of the camera as a promising method for IACT full-event reconstruction and present the performance of the method on observational data using CTLearn, a package for IACT event reconstruction that exploits deep learning. arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2211.16009v1

magic deep performance la palma canary islands telescopes arxiv gev cherenkov convolutional neural networks iact

The performance of the MAGIC telescopes using deep convolutional neural networks with CTLearn

Play Episode Listen Later Nov 29, 2022 0:39

The performance of the MAGIC telescopes using deep convolutional neural networks with CTLearn by T. Miener et al. on Tuesday 29 November The Major Atmospheric Gamma Imaging Cherenkov (MAGIC) telescope system is located on the Canary Island of La Palma and inspects the very high-energy (VHE, few tens of GeV and above) gamma-ray sky. MAGIC consists of two imaging atmospheric Cherenkov telescopes (IACTs), which capture images of the air showers originating from the absorption of gamma rays and cosmic rays by the atmosphere, through the detection of Cherenkov photons emitted in the shower. The sensitivity of IACTs to gamma-ray sources is mainly determined by the ability to reconstruct the properties (type, energy, and arrival direction) of the primary particle generating the air shower. The state-of-the-art IACT pipeline for shower reconstruction is based on the parameterization of the shower images by extracting geometric and stereoscopic features and machine learning algorithms like random forest or boosted decision trees. In this contribution, we explore deep convolutional neural networks applied directly to the pixelized images of the camera as a promising method for IACT full-event reconstruction and present the performance of the method on observational data using CTLearn, a package for IACT event reconstruction that exploits deep learning. arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2211.16009v1

magic deep performance la palma canary islands telescopes arxiv gev cherenkov convolutional neural networks iact

Application of Convolutional Neural Networks to Predict Magnetic Fields Directions in Turbulent Clouds

3d application clouds predict directions taurus turbulent magnetic fields planck arxiv convolutional neural networks

Play Episode Listen Later Nov 29, 2022 0:29

Application of Convolutional Neural Networks to Predict Magnetic Fields Directions in Turbulent Clouds by Duo Xu et al. on Tuesday 29 November We adopt the deep learning method CASI-3D (Convolutional Approach to Structure Identification-3D) to infer the orientation of magnetic fields in sub-/trans- Alfvenic turbulent clouds from molecular line emission. We carry out magnetohydrodynamic simulations with different magnetic field strengths and use these to generate synthetic observations. We apply the 3D radiation transfer code RADMC-3d to model 12CO and 13CO (J = 1-0) line emission from the simulated clouds and then train a CASI-3D model on these line emission data cubes to predict magnetic field morphology at the pixel level. The trained CASI-3D model is able to infer magnetic field directions with low error (< 10deg for sub-Alfvenic samples and

Application of Convolutional Neural Networks to Predict Magnetic Fields Directions in Turbulent Clouds

3d application clouds predict directions taurus turbulent magnetic fields planck arxiv convolutional neural networks

Play Episode Listen Later Nov 28, 2022 0:30

Application of Convolutional Neural Networks to Predict Magnetic Fields Directions in Turbulent Clouds by Duo Xu et al. on Monday 28 November We adopt the deep learning method CASI-3D (Convolutional Approach to Structure Identification-3D) to infer the orientation of magnetic fields in sub-/trans- Alfvenic turbulent clouds from molecular line emission. We carry out magnetohydrodynamic simulations with different magnetic field strengths and use these to generate synthetic observations. We apply the 3D radiation transfer code RADMC-3d to model 12CO and 13CO (J = 1-0) line emission from the simulated clouds and then train a CASI-3D model on these line emission data cubes to predict magnetic field morphology at the pixel level. The trained CASI-3D model is able to infer magnetic field directions with low error (< 10deg for sub-Alfvenic samples and

Application of Convolutional Neural Networks to Predict Magnetic Fields Directions in Turbulent Clouds

3d application clouds predict directions taurus turbulent magnetic fields planck arxiv convolutional neural networks

Play Episode Listen Later Nov 28, 2022 0:32

Using deep convolutional neural networks to test why human face recognition works the way it does

Play Episode Listen Later Nov 24, 2022

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2022.11.23.517478v1?rss=1 Authors: Dobs, K., Yuan, J., Martinez, J., Kanwisher, N. Abstract: Human face recognition is highly accurate, and exhibits a number of distinctive and well documented behavioral "signatures" such as the use of a characteristic representational space, the disproportionate performance cost when stimuli are presented upside down, and the drop in accuracy for faces from races the participant is less familiar with. These and other phenomena have long been taken as evidence that face recognition is "special". But why does human face perception exhibit these properties in the first place? Here we use convolutional neural networks (CNNs) to test the hypothesis that all of these signatures of human face perception result from optimization for the task of face recognition. Indeed, as predicted by this hypothesis, these phenomena are all found in CNNs trained on face recognition, but not in CNNs trained on object recognition, even when the amount of face experience is matched. To test whether these signatures are in principle specific to faces, we optimized a CNN on car discrimination and tested it on upright and inverted car images. As for face perception, the car-trained network showed a drop in performance for inverted versus upright cars. Similarly, CNNs trained only on inverted faces produce an inverted inversion effect. These findings show that the behavioral signatures of human face perception reflect and are well explained as the result of optimization for the task of face recognition, and that the nature of the computations underlying this task may not be so "special" after all. Copy rights belong to original authors. Visit the link for more info Podcast created by Paper Player, LLC

deep cnn llc martinez copy yuan face recognition human face biorxiv convolutional neural networks abstract human

Application of convolutional neural networks for data analysis in TAIGA-HiSCORE experiment

Play Episode Listen Later Nov 23, 2022 0:41

Application of convolutional neural networks for data analysis in TAIGA-HiSCORE experiment by Anna Vlaskina et al. on Wednesday 23 November The TAIGA experimental complex is a hybrid observatory for high-energy gamma-ray astronomy in the range from 10 TeV to several EeV. The complex consists of such installations as TAIGA- IACT, TAIGA-HiSCORE and a number of others. The TAIGA-HiSCORE facility is a set of wide-angle synchronized stations that detect Cherenkov radiation scattered over a large area. TAIGA-HiSCORE data provides an opportunity to reconstruct shower characteristics, such as shower energy, direction of arrival, and axis coordinates. The main idea of the work is to apply convolutional neural networks to analyze HiSCORE events, considering them as images. The distribution of registration times and amplitudes of events recorded by HiSCORE stations is used as input data. The paper presents the results of using convolutional neural networks to determine the characteristics of air showers. It is shown that even a simple model of convolutional neural network provides the accuracy of recovering EAS parameters comparable to the traditional method. Preliminary results of air shower parameters reconstruction obtained in a real experiment and their comparison with the results of traditional analysis are presented. arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2211.12197v1

experiments application preliminary data analysis eas taiga tev arxiv cherenkov convolutional neural networks eev

Application of convolutional neural networks for data analysis in TAIGA-HiSCORE experiment

Play Episode Listen Later Nov 22, 2022 0:40

Application of convolutional neural networks for data analysis in TAIGA-HiSCORE experiment by Anna Vlaskina et al. on Tuesday 22 November The TAIGA experimental complex is a hybrid observatory for high-energy gamma-ray astronomy in the range from 10 TeV to several EeV. The complex consists of such installations as TAIGA- IACT, TAIGA-HiSCORE and a number of others. The TAIGA-HiSCORE facility is a set of wide-angle synchronized stations that detect Cherenkov radiation scattered over a large area. TAIGA-HiSCORE data provides an opportunity to reconstruct shower characteristics, such as shower energy, direction of arrival, and axis coordinates. The main idea of the work is to apply convolutional neural networks to analyze HiSCORE events, considering them as images. The distribution of registration times and amplitudes of events recorded by HiSCORE stations is used as input data. The paper presents the results of using convolutional neural networks to determine the characteristics of air showers. It is shown that even a simple model of convolutional neural network provides the accuracy of recovering EAS parameters comparable to the traditional method. Preliminary results of air shower parameters reconstruction obtained in a real experiment and their comparison with the results of traditional analysis are presented. arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2211.12197v1

experiments application preliminary data analysis eas taiga tev arxiv cherenkov convolutional neural networks eev

Emergence of perceptual reorganisation from prior knowledge in human development and Convolutional Neural Networks

young llc copy emergence mclean human development dekker zheng groen lisi cornet perceptual underpinning reorganisation biorxiv convolutional neural networks

Play Episode Listen Later Nov 22, 2022

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2022.11.21.517321v1?rss=1 Authors: Milne, G. A., Lisi, M., McLean, A., Zheng, R., Groen, I. I. A., Dekker, T. M. Abstract: The use of prior knowledge to guide perception is fundamental to human vision, especially under challenging viewing circumstances. Underpinning current theories of predictive coding, prior knowledge delivered to early sensory areas via cortical feedback connections can reshape perception of ambiguous stimuli, such as 'two-tone' images. Despite extensive interest and ongoing research into this process of perceptual reorganisation in the adult brain, it is not yet fully understood how or when the efficient use of prior knowledge for visual perception develops. Here we show for the first time that adult-like levels of perceptual reorganisation do not emerge until late childhood. We used a behavioural two-tone paradigm to isolate the effects of prior knowledge on visual perception in children aged 4 - 12 years and adults, and found a clear developmental progression in the perceptual benefit gained from informative cues. Whilst photo cueing reliably triggered perceptual reorganisation of two-tones for adults, 4- to 9-year-olds' performed significantly poorer immediately after cueing than within-subject benchmarks of recognition. Young childens' behaviour revealed perceptual biases towards local image features, as has been seen in image classification neural networks. We tested three such models (AlexNet, CorNet and NASNet) on two-tone classification, and while we found that network depth and recurrence may improve recognition, the best-performing network behaved similarly to young children. Our results reveal a prolonged development of prior-knowledge-guided vision throughout childhood, a process which may be central to other perceptual abilities that continue developing throughout childhood. This highlights the importance of effective reconciliation of signal and prediction for robust perception in both human and computational vision systems. Copy rights belong to original authors. Visit the link for more info Podcast created by Paper Player, LLC

Modelling surface color discrimination under different lighting environments using image chromatic statistics and convolutional neural networks

cnn llc color statistics guided copy discrimination surface lighting environments modelling chromatics smithson morimoto biorxiv convolutional neural networks

Play Episode Listen Later Nov 3, 2022

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2022.11.02.514864v1?rss=1 Authors: Ponting, S., Morimoto, T., Smithson, H. E. Abstract: We modeled discrimination thresholds for object colors under different lighting environments. Firstly we built models based on chromatic statistics, testing 60 models in total. Secondly we trained convolutional neural networks (CNNs), using 160,280 images labeled either by the ground-truth or by human responses. No single chromatic statistics model was sufficient to describe human discrimination thresholds across conditions, while human-response-trained CNNs nearly perfectly predicted human thresholds. Guided by region-of-interest analysis of the network, we modified the chromatic statistics models to use only the lower regions of the objects, which substantially improved performance. Copy rights belong to original authors. Visit the link for more info Podcast created by Paper Player, LLC

Determine the Core Structure and Nuclear Equation of State of Rotating Core-Collapse Supernovae with Gravitational Waves by Convolutional Neural Networks

state structure collapse nuclear determine pan equation eos rotating gravitational waves arxiv supernovae convolutional neural networks

Play Episode Listen Later Sep 21, 2022 0:50

Determine the Core Structure and Nuclear Equation of State of Rotating Core-Collapse Supernovae with Gravitational Waves by Convolutional Neural Networks by Yang-Sheng Chao et al. on Wednesday 21 September Detecting gravitational waves from a nearby core-collapse supernova would place meaningful constraints on the supernova engine and nuclear equation of state. Here we use Convolutional Neural Network models to identify the core rotational rates, rotation length scales, and the nuclear equation of state (EoS), using the 1824 waveforms from Richers et al. (2017) for a 12 solar mass progenitor. High prediction accuracy for the classifications of the rotation length scales ($93%$) and the rotational rates ($95%$) can be achieved using the gravitational wave signals from -10 ms to 6 ms core bounce. By including additional 48 ms signals during the prompt convection phase, we could achieve $96%$ accuracy on the classification of four major EoS groups. Combining three models above, we could correctly predict the core rotational rates, rotation length scales, and the EoS at the same time with more than $85%$ accuracy. Finally, applying a transfer learning method for additional 74 waveforms from FLASH simulations (Pan et al. 2018), we show that our model using Richers' waveforms could successfully predict the rotational rates from Pan's waveforms even for a continuous value with a mean absolute errors of 0.32 rad s$^{-1}$ only. These results demonstrate a much broader parameter regimes our model can be applied for the identification of core-collapse supernova events through GW signals. arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2209.10089v1

Convolutional Neural Networks for Searching Superflares from Pixel-level Data of the Transiting Exoplanet Survey Satellite

Play Episode Listen Later Sep 18, 2022 0:13

Convolutional Neural Networks for Searching Superflares from Pixel-level Data of the Transiting Exoplanet Survey Satellite by Zuo-Lin Tu et al. on Sunday 18 September In this work, six convolutional neural networks (CNNs) have been trained based on %different feature images and arrays from the database including 15,638 superflare candidates on solar-type stars, which are collected from the three-years observations of Transiting Exoplanet Survey Satellite ({em TESS}). These networks are used to replace the artificially visual inspection, which was a direct way to search for superflares, and exclude false positive events in recent years. Unlike other methods, which only used stellar light curves to search superflare signals, we try to identify superflares through {em TESS} pixel-level data with lower risks of mixing false positive events, and give more reliable identification results for statistical analysis. The evaluated accuracy of each network is around 95.57%. After applying ensemble learning to these networks, stacking method promotes accuracy to 97.62% with 100% classification rate, and voting method promotes accuracy to 99.42% with relatively lower classification rate at 92.19%. We find that superflare candidates with short duration and low peak amplitude have lower identification precision, as their superflare-features are hard to be identified. The database including 71,732 solar-type stars and 15,638 superflare candidates from {em TESS} with corresponding feature images and arrays, and trained CNNs in this work are public available. arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2204.04019v2

data cnn searching satellites pixel exoplanets arxiv convolutional neural networks transiting exoplanet survey satellite

Lessons Learned from the Two Largest Galaxy Morphological Classification Catalogues built by Convolutional Neural Networks

cnn built lessons learned galaxy largest classification arxiv morphological dark energy survey convolutional neural networks

Play Episode Listen Later Sep 15, 2022 0:41

Lessons Learned from the Two Largest Galaxy Morphological Classification Catalogues built by Convolutional Neural Networks by Ting-Yun Cheng et al. on Thursday 15 September We compare the two largest galaxy morphology catalogues, which separate early and late type galaxies at intermediate redshift. The two catalogues were built by applying supervised deep learning (convolutional neural networks, CNNs) to the Dark Energy Survey data down to a magnitude limit of $sim$21 mag. The methodologies used for the construction of the catalogues include differences such as the cutout sizes, the labels used for training, and the input to the CNN - monochromatic images versus $gri$-band normalized images. In addition, one catalogue is trained using bright galaxies observed with DES ($i

Joint optimization of wavefront sensing and reconstruction with automatic differentiation

joint reconstruction optimization automatic subsequently differentiation sensing fourier arxiv wavefront convolutional neural networks

Play Episode Listen Later Sep 13, 2022 0:49

Joint optimization of wavefront sensing and reconstruction with automatic differentiation by Rico Landman et al. on Tuesday 13 September High-contrast imaging instruments need extreme wavefront control to directly image exoplanets. This requires highly sensitive wavefront sensors which optimally make use of the available photons to sense the wavefront. Here, we propose to numerically optimize Fourier-filtering wavefront sensors using automatic differentiation. First, we optimize the sensitivity of the wavefront sensor for different apertures and wavefront distributions. We find sensors that are more sensitive than currently used sensors and close to the theoretical limit, under the assumption of monochromatic light. Subsequently, we directly minimize the residual wavefront error by jointly optimizing the sensing and reconstruction. This is done by connecting differentiable models of the wavefront sensor and reconstructor and alternatingly improving them using a gradient-based optimizer. We also allow for nonlinearities in the wavefront reconstruction using Convolutional Neural Networks, which extends the design space of the wavefront sensor. Our results show that optimization can lead to wavefront sensors that have improved performance over currently used wavefront sensors. The proposed approach is flexible, and can in principle be used for any wavefront sensor architecture with free design parameters. arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2209.05904v1

Harvesting the Lyα forest with convolutional neural networks

cnn delta harvesting sn voigt ly arxiv convolutional neural networks

Play Episode Listen Later Sep 6, 2022 0:12

Harvesting the Lyα forest with convolutional neural networks by Ting-Yun Cheng et al. on Tuesday 06 September We develop a machine learning based algorithm using a convolutional neural network (CNN) to identify low HI column density Ly$alpha$ absorption systems ($log{N_{mathrm{HI}}}/{rm cm}^{-2}

Small Training Dataset Convolutional Neural Networks for Application Specific Super-Resolution Microscopy

training resolutions cnn application copy ml db demonstrate microscopy dataset biorxiv convolutional neural networks

Play Episode Listen Later Aug 29, 2022

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2022.08.29.505633v1?rss=1 Authors: Mannam, V., Howard, S. Abstract: Machine learning (ML) models based on convolutional neural networks (CNNs) have been used to significantly increase microscopy resolution, speed (signal-to-noise ratio), and data interpretation. The bottleneck in developing effective ML systems is often the need to acquire large datasets to train the neural network. This paper demonstrates how adding a dense encoder-decoder block can be useful to effectively train a CNN that provides super-resolution images from conventional diffraction-limited microscopy images when trained using a small dataset containing 15 field-of-views (FOVs). DenseED blocks use a dense layer that concatenates features from the previous convolutional layer to the next convolutional layer. Demonstrate using DenseED blocks in fully convolutional networks (FCNs) to estimate the super-resolution images when trained with a small training dataset (15 FOVs) of human cells from the Widefield2SIM dataset and the fluorescent-labeled fixed bovine pulmonary artery endothelial cells (BPAE samples). Conventional ML models without DenseED blocks trained on small datasets fail to accurately estimate super-resolution images while models, including the DenseED blocks can. The average resolution and peak signal-to-noise ratio (PSNR) improvements achieved using DenseED blocks in FCNs when trained with 15 FOVs are 2 times and ~3.2 dB, respectively. In addition, we evaluated various configurations of target image generation methods (experimentally captured target and computationally generated target) that are used to train the FCNs with and without DenseED blocks and showed with DenseED blocks outperform compared to simple FCNs without DenseED blocks. Hence, the proposed approach indicates that microscopy applications can use DenseED blocks to train on smaller datasets that are application specific/experimental modality specific imaging platforms such as MRI/X-ray and other in vivo imaging modalities. Copy rights belong to original authors. Visit the link for more info Podcast created by PaperPlayer

Language independence and material properties

Underrated ML

Play Episode Listen Later Jul 26, 2022 94:05

This week we are joined by Sebastian Ruder. He is a research scientist at DeepMind, London. He has also worked at a variety of institutions such as AYLIEN, Microsoft, IBM's Extreme Blue, Google Summer of Code, and SAP. These experiences were completed in tangent with his studies which included studying Computational Linguistics at the University of Heidelberg, Germany and at Trinity College, Dublin before undertaking a PhD in Natural Language Processing and Deep Learning at the Insight Research Centre for Data Analytics.This week we discuss language independence and diversity in natural language processing whilst also taking a look at the attempts to identify material properties from images.As discussed in the podcast if you would like to donate to the current campaign of "CREATE DONATE EDUCATE" which supports Stop Hate UK then please find the link below:https://www.shorturl.at/glmszPlease also find additional links to help support black colleagues in the area of research;Black in AI twitter account: https://twitter.com/black_in_aiMentoring and proofreading sign-up to support our Black colleagues in research: https://twitter.com/le_roux_nicolas/status/1267896907621433344?s=20Underrated ML Twitter: https://twitter.com/underrated_mlSebastian Ruder Twitter: https://twitter.com/seb_ruderPlease let us know who you thought presented the most underrated paper in the form below: https://forms.gle/97MgHvTkXgdB41TC8Links to the papers:“On Achieving and Evaluating Language-Independence in NLP” - https://journals.linguisticsociety.org/elanguage/lilt/article/view/2624.html"The State and Fate of Linguistic Diversity and Inclusion in the NLP World” - https://arxiv.org/abs/2004.09095"Recognizing Material Properties from Images" - https://arxiv.org/pdf/1801.03127.pdfAdditional Links:Student perspectives on applying to NLP PhD programs: https://blog.nelsonliu.me/2019/10/24/student-perspectives-on-applying-to-nlp-phd-programs/Tim Dettmer's post on how to pick your grad school: https://timdettmers.com/2020/03/10/how-to-pick-your-grad-school/Rachel Thomas' blog post on why you should blog: https://medium.com/@racheltho/why-you-yes-you-should-blog-7d2544ac1045Emily Bender's The Gradient article: https://thegradient.pub/the-benderrule-on-naming-the-languages-we-study-and-why-it-matters/Paper on order-sensitive vs order-free methods: https://www.aclweb.org/anthology/N19-1253.pdf"Exploring the Origins and Prevalence of Texture Bias in Convolutional Neural Networks": https://arxiv.org/abs/1911.09071Sebastian's website where you can find all his blog posts: https://ruder.io/

The National Science Foundation and Advancement in Artificial Intelligence

Mind Matters

Play Episode Listen Later Apr 28, 2022 115:04

Early in his career, IEEE fellow and retired National Science Foundation program director Paul Werbos developed the neural network training algorithm known as error backpropagation, which has been foundational to the vast majority of today’s advances in artificial intelligence. Listen in as he discusses his work in this area and other topics, including his tenure with the National Science Foundation,… Source

artificial intelligence albert einstein multiverse big tech advancement deep learning national science foundation alan turing ieee geoffrey hinton machine intelligence david deutsch john wheeler marvin minsky norbert wiener mind body problem artificial neural networks backpropagation convolutional neural networks

dl convolutional neural networks

my way on medicine

Play Episode Listen Later Dec 22, 2021 6:58

convolutional neural networks

MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications

Papers Read on AI

Play Episode Listen Later Sep 11, 2021 22:06

Classic: We present a class of efficient models called MobileNets for mobile and embedded vision applications. MobileNets are based on a streamlined architecture that uses depth-wise separable convolutions to build light weight deep neural networks. We introduce two simple global hyper-parameters that efficiently trade off between latency and accuracy. These hyper-parameters allow the model builder to choose the right sized model for their application based on the constraints of the problem. We present extensive experiments on resource and accuracy tradeoffs and show strong performance compared to other popular models on ImageNet classification. 2017: Andrew G. Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, Tobias Weyand, M. Andreetto, Hartwig Adam https://arxiv.org/pdf/1704.04861.pdf

vision mobile applications efficient imagenet andrew g convolutional neural networks

#3 - Artificial intelligence for images. Understanding Convolutional Neural Networks - CNNs.

Life with AI

Play Episode Listen Later Jun 17, 2021 17:14

In this episode I talk about artificial intelligence for image. I explain the idea behind one of the most popular AI algorithm, the neural networks and show how it's able to learn.

ai artificial intelligence images convolutional neural networks

Paul Werbos: The National Science Foundation and AI

Mind Matters

Play Episode Listen Later Jun 10, 2021 26:39

In today’s episode, Dr. Robert J. Marks continues his conversation with Dr. Paul Werbos, the inventor of the most commonly used technique to train artificial neural networks. Listen in as they turn to the National Science Foundation, its role in steering research in artificial intelligence, and the major turning points in machine intelligence that Dr. Werbos witnessed as a program… Source

artificial intelligence deep learning national science foundation geoffrey hinton machine intelligence convolutional neural networks robert j marks

Paul Werbos: The National Science Foundation and AI

Mind Matters

Play Episode Listen Later Jun 10, 2021 26:39

artificial intelligence deep learning national science foundation geoffrey hinton machine intelligence convolutional neural networks robert j marks

MLP-Mixer: An all-MLP Architecture for Vision (Machine Learning Research Paper Explained)

Yannic Kilcher Videos (Audio Only)

Play Episode Listen Later May 10, 2021 28:11

#mixer #google #imagenet Convolutional Neural Networks have dominated computer vision for nearly 10 years, and that might finally come to an end. First, Vision Transformers (ViT) have shown remarkable performance, and now even simple MLP-based models reach competitive accuracy, as long as sufficient data is used for pre-training. This paper presents MLP-Mixer, using MLPs in a particular weight-sharing arrangement to achieve a competitive, high-throughput model and it raises some interesting questions about the nature of learning and inductive biases and their interaction with scale for future research. OUTLINE: 0:00 - Intro & Overview 2:20 - MLP-Mixer Architecture 13:20 - Experimental Results 17:30 - Effects of Scale 24:30 - Learned Weights Visualization 27:25 - Comments & Conclusion Paper: https://arxiv.org/abs/2105.01601 Abstract: Convolutional Neural Networks (CNNs) are the go-to model for computer vision. Recently, attention-based networks, such as the Vision Transformer, have also become popular. In this paper we show that while convolutions and attention are both sufficient for good performance, neither of them are necessary. We present MLP-Mixer, an architecture based exclusively on multi-layer perceptrons (MLPs). MLP-Mixer contains two types of layers: one with MLPs applied independently to image patches (i.e. "mixing" the per-location features), and one with MLPs applied across patches (i.e. "mixing" spatial information). When trained on large datasets, or with modern regularization schemes, MLP-Mixer attains competitive scores on image classification benchmarks, with pre-training and inference cost comparable to state-of-the-art models. We hope that these results spark further research beyond the realms of well established CNNs and Transformers. Authors: Ilya Tolstikhin, Neil Houlsby, Alexander Kolesnikov, Lucas Beyer, Xiaohua Zhai, Thomas Unterthiner, Jessica Yung, Daniel Keysers, Jakob Uszkoreit, Mario Lucic, Alexey Dosovitskiy ERRATA: Here is their definition of what the 5-shot classifier is: "we report the few-shot accuracies obtained by solving the L2-regularized linear regression problem between the frozen learned representations of images and the labels" Links: TabNine Code Completion (Referral): http://bit.ly/tabnine-yannick YouTube: https://www.youtube.com/c/yannickilcher Twitter: https://twitter.com/ykilcher Discord: https://discord.gg/4H8xxDF BitChute: https://www.bitchute.com/channel/yann... Minds: https://www.minds.com/ykilcher Parler: https://parler.com/profile/YannicKilcher LinkedIn: https://www.linkedin.com/in/yannic-ki... BiliBili: https://space.bilibili.com/1824646584 If you want to support me, the best thing to do is to share out the content :) If you want to support me financially (completely optional and voluntary, but a lot of people have asked for this): SubscribeStar: https://www.subscribestar.com/yannick... Patreon: https://www.patreon.com/yannickilcher Bitcoin (BTC): bc1q49lsw3q325tr58ygf8sudx2dqfguclvngvy2cq Ethereum (ETH): 0x7ad3513E3B8f66799f507Aa7874b1B0eBC7F85e2 Litecoin (LTC): LQW2TRyKYetVC8WjFkhpPhtpbDM4Vw7r9m Monero (XMR): 4ACL8AGrEo5hAir8A9CeVrW8pEauWvnp1WnSDZxW7tziCDLhZAGsgzhRQABDnFy8yuM9fWJDviJPHKRjV4FWt19CJZN9D4n

vision cnn discord effects minds architecture transformers machine learning parler mixer outline l2 mlp research papers bilibili mlps learning research convolutional neural networks

Episode 07: Yujia Huang, Caltech, on neuro-inspired generative models

Generally Intelligent

Play Episode Listen Later Mar 18, 2021 65:42

Yujia Huang (@YujiaHuangC) is a PhD student at Caltech, working at the intersection of deep learning and neuroscience. She worked on optics and biophotonics before venturing into machine learning. Now, she hopes to design “less artificial” artificial intelligence. Her most recent paper at NeurIPS is Neural Networks with Recurrent Generative Feedback, introducing Convolutional Neural Networks with Feedback (CNN-F). Yujia is open to working with collaborators from many areas: neuroscience, signal processing, and control experts — in addition to those interested in generative models for classification. Feel free to reach out to her! Highlights from our conversation:

ai phd tips adapting neuro huang turing caltech neural networks neurips generative models convolutional neural networks

Episode 29: ML 023: Inside Machine Learning with Edward Raff

Adventures in Machine Learning

Play Episode Listen Later Mar 16, 2021 58:34

We have a new panelist! Plus, Edward Raff joins the Adventure to discuss his new book Inside Machine Learning. He walks us through Convolutional Neural Networks and then talks us through to build, train, and use them to solve problems through Machine Learning. The conversation ranges into having good data sets, tweaking your network, and when a Convolutional Neural Network is an appropriate tool for the problem you're trying to solve. Panel Charles Max Wood Miguel Morales Guest Edward Raff Sponsors Dev Heroes Accelerator Picks Charles- Docker Charles- Paramount+ Edward- Java | Oracle Edward- Make-A-Wish America Miguel- Full Stack Deep Learning - Spring 2021

adventure panel machine learning paramount docker raff devchat miguel morales charles max wood convolutional neural networks

Visual Illusions Deceiving Neural Networks

Data Skeptic

Play Episode Listen Later Jan 1, 2021 33:43

Today on the show we have Adrian Martin, a Postdoctorial researcher from the Univeristy of Pompeu Fabra in Barcelona, Spain. He comes on the show today to discuss his research from the paper “Convolutional Neural Networks can be Decieved by Visual Illusions.” Workes Mentioned in Paper: “Convolutional Neural Networks can be Decieved by Visual Illusions.” by Alexander Gomez-Villa, Adrian Martin, Javier Vazquez-Corral, and Marcelo Bertalmio Examples: Snake Illusions https://www.illusionsindex.org/i/rotating-snakes Twitter: Alex: @alviur Adrian: @adriMartin13

spain barcelona visual illusions deceiving neural networks univeristy adrian martin convolutional neural networks

High-Throughput Image-Based Plant Stand Count Estimation Using Convolutional Neural Networks

Play Episode Listen Later Nov 6, 2020

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.11.05.370437v1?rss=1 Authors: Khaki, S., Pham, H., Han, Y., Kent, W., Wang, L. Abstract: The future landscape of modern farming and plant breeding is rapidly changing due to the complex needs of our society. The explosion of collectable data has started a revolution in agriculture to the point where innovation must occur. To a commercial organization, the accurate and efficient collection of information is necessary to ensure that optimal decisions are made at key points of the breeding cycle. However, due to the sheer size of a breeding program and current resource limitations, the ability to collect precise data on individual plants is not possible. In particular, efficient phenotyping of crops to record its color, shape, chemical properties, disease susceptibility, etc. is severely limited due to labor requirements and, oftentimes, expert domain knowledge. In this paper, we propose a deep learning based approach, named DeepStand, for image-based corn stand counting at early phenological stages. The proposed method adopts a truncated VGG-16 network as a backbone feature extractor and merges multiple feature maps with different scales to make the network robust against scale variation. Our extensive computational experiments suggest that our proposed method can successfully count corn stands and out-perform other state-of-the-art methods. It is the goal of our work to be used by the larger agricultural community as a way to enable high-throughput phenotyping without the use of extensive time and labor requirements. Copy rights belong to original authors. Visit the link for more info

stand plant kent copy wang pham estimation throughput biorxiv convolutional neural networks vgg

fastISM: Performant in-silico saturation mutagenesis for convolutional neural networks

copy deep sea saturation ism performant biorxiv mutagenesis convolutional neural networks

Play Episode Listen Later Oct 13, 2020

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.10.13.337147v1?rss=1 Authors: Nair, S., Shrikumar, A., Kundaje, A. Abstract: Deep learning models such as convolutional neural networks are able to accurately map biological sequences to associated functional readouts and properties by learning predictive de novo representations. In-silico saturation mutagenesis (ISM) is a popular feature attribution technique for inferring contributions of all characters in an input sequence to the model's predicted output. The main drawback of ISM is its runtime, as it involves multiple forward propagations of all possible mutations of each character in the input sequence through the trained model to predict the effects on the output. We present fastISM, an algorithm that speeds up ISM by a factor of over 10x for commonly used convolutional neural network architectures. fastISM is based on the observations that the majority of computation in ISM is spent in convolutional layers, and a single mutation only disrupts a limited region of intermediate layers, rendering most computation redundant. fastISM reduces the gap between backpropagation-based feature attribution methods and ISM. It far surpasses the runtime of backpropagation-based methods on multi-output architectures, making it feasible to run ISM on a large number of sequences. An easy-to-use Keras/TensorFlow 2 implementation of fastISM is available at https://github.com/kundajelab/fastISM, and a hands-on tutorial at https://colab.research.google.com/github/kundajelab/fastISM/blob/master/notebooks/colab/DeepSEA.ipynb. Copy rights belong to original authors. Visit the link for more info

DEELIG: A Deep Learning-based approach to predict protein-ligand binding affinity

copy machine learning protein predict mam binding deep learning affinity biorxiv ligand convolutional neural networks

Play Episode Listen Later Sep 28, 2020

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.09.28.316224v1?rss=1 Authors: Ahmed, A., Mam, B., Sowdhamini, R. Abstract: Protein-ligand binding prediction has extensive biological significance. Binding affinity helps in understanding the degree of protein-ligand interactions and has wide protein applications. Protein-ligand docking using virtual screening and molecular dynamic simulations are required to predict the binding affinity of a ligand to its cognate receptor. In order to perform such analyses, it requires intense computational power and it becomes impossible to cover the entire chemical space of small molecules. It has been aided by a shift towards using Machine Learning-based methodologies that aids in binding prediction using regression. Recent developments using deep learning has enabled us to make sense of massive amounts of complex datasets. Herein, the ability of the model to learn intrinsic patterns in a complex plane of data is the strength of the approach. Here, we have incorporated Convolutional Neural Networks that find spatial relationships among data to help us predict affinity of binding of proteins in whole superfamilies towards a diverse set of ligands. The models were trained and validated using a detailed methodology for feature extraction. We have also tested DEELIG on protein complexes relevant to the current public health scenario. Our approach to network construction and training on protein-ligand dataset prepared in-house has provided significantly better results than previously existing methods in the field. Copy rights belong to original authors. Visit the link for more info

CancerSiamese: one-shot learning for primary and metastatic tumor classification

Play Episode Listen Later Sep 9, 2020

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.09.07.286583v1?rss=1 Authors: Mostavi, M., Chiu, Y.-C., Chen, Y., Huang, Y. Abstract: We consider cancer classification based on one single gene expression profile. We proposed CancerSiamese, a new one-shot learning model, to predict the cancer type of a query primary or metastatic tumor sample based on a support set that contains only one known sample for each cancer type. CancerSiamese receives pairs of gene expression profiles and learns a representation of similar or dissimilar cancer types through two parallel Convolutional Neural Networks joined by a similarity function. We trained CancerSiamese for both primary and metastatic cancer type predictions using samples from TCGA and MET500. Test results for different N-way predictions yielded an average accuracy improvement of 8% and 4% over the benchmark 1-Nearest Neighbor (1-NN) classifier for primary and metastatic tumors, respectively. Moreover, we applied the guided gradient saliency map and feature selection to CancerSiamese to identify and analyze the marker-gene candidates for primary and metastatic cancers. Our work demonstrated, for the first time, the feasibility of applying one-shot learning for expression-based cancer type prediction when gene expression data of cancer types are limited and could inspire new and ingenious applications of one-shot and few-shot learning solutions for improving cancer diagnosis, treatment planning, and our understanding of cancer. Copy rights belong to original authors. Visit the link for more info

learning primary copy chen tumors one shot huang classification chiu nn metastatic biorxiv convolutional neural networks tcga

An automated framework for efficiently designing deep convolutional neural networks in genomics

deep cnn park designing framework nas copy interpretation tuning automated genomics efficiently biorxiv convolutional neural networks

Play Episode Listen Later Aug 19, 2020

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.08.18.251561v1?rss=1 Authors: Zhang, Z., Park, C. Y., Theesfeld, C. L., Troyanskaya, O. G. Abstract: Convolutional neural networks (CNNs) have become a standard for analysis of biological sequences. Tuning of network architectures is essential for CNN's performance, yet it requires substantial knowledge of machine learning and commitment of time and effort. This process thus imposes a major barrier to broad and effective application of modern deep learning in genomics. Here, we present AMBER, a fully automated framework to efficiently design and apply CNNs for genomic sequences. AMBER designs optimal models for user-specified biological questions through the state-of-the-art Neural Architecture Search (NAS). We applied AMBER to the task of modelling genomic regulatory features and demonstrated that the predictions of the AMBER-designed model are significantly more accurate than the equivalent baseline non-NAS models and match or even exceed published expert-designed models. Interpretation of AMBER architecture search revealed its design principles of utilizing the full space of computational operations for accurately modelling genomic sequences. Furthermore, we illustrated the use of AMBER to accurately discover functional genomic variants in allele-specific binding and disease heritability enrichment. AMBER provides an efficient automated method for designing accurate deep learning models in genomics. Copy rights belong to original authors. Visit the link for more info

Enhancing SNR and generating contrast for cryo-EM images with convolutional neural networks

PaperPlayer biorxiv biophysics

Play Episode Listen Later Aug 16, 2020

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.08.16.253070v1?rss=1 Authors: Palovcak, E., Asarnow, D., Campbell, M. G., Yu, Z., Cheng, Y. Abstract: In cryogenic electron microscopy (cryo-EM) of radiation-sensitive biological samples, both the signal-to-noise ratio (SNR) and the contrast of images are critically important in the image processing pipeline. Classic methods improve low-frequency image contrast experimentally, by imaging with high defocus, or computationally, by applying various types of low-pass filter. These contrast improvements typically come at the expense of high-frequency SNR, which is suppressed by high-defocus imaging and removed by low pass filtration. Here, we demonstrate that a convolutional neural network (CNN) denoising algorithm can be used to significantly enhance SNR and generate contrast in cryo-EM images. We provide a quantitative evaluation of bias introduced by the denoising procedure and its influences on image processing and three-dimensional reconstructions. Our study suggests that besides enhancing the visual contrast of cryo-EM images, the enhanced SNR of denoised images may facilitate better outcomes in the other parts of the image processing pipeline, such as classification and 3D alignment. Overall, our results provide a ground of using denoising CNNs in the cryo-EM image processing pipeline. Copy rights belong to original authors. Visit the link for more info

3d cnn campbell classic images copy enhancing generating contrast cheng yu snr biorxiv convolutional neural networks cryo em

The relative coding strength of object identity and nonidentity features in human occipito-temporal cortex and convolutional neural networks

identity strength cnn similar copy coding object relative temporal cortex euclidean biorxiv convolutional neural networks

Play Episode Listen Later Aug 12, 2020

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.08.11.246967v1?rss=1 Authors: Xu, Y., Vaziri-Pashkam, M. Abstract: Any given visual object input is characterized by multiple visual features, such as identity, position and size. Despite the usefulness of identity and nonidentity features in vision and their joint coding throughout the primate ventral visual processing pathway, they have so far been studied relatively independently. Here we document the relative coding strength of object identity and nonidentity features in a brain region and how this may change across the human ventral visual pathway. We examined a total of four nonidentity features, including two Euclidean features (position and size) and two non-Euclidean features (image statistics and spatial frequency content of an image). Overall, identity representation increased and nonidentity feature representation decreased along the ventral visual pathway, with identity outweighed the non-Euclidean features, but not the Euclidean ones, in higher levels of visual processing. A similar analysis was performed in 14 convolutional neural networks (CNNs) pretrained to perform object categorization with varying architecture, depth, and with/without recurrent processing. While the relative coding strength of object identity and nonidentity features in lower CNN layers matched well with that in early human visual areas, the match between higher CNN layers and higher human visual regions were limited. Similar results were obtained regardless of whether a CNN was trained with real-world or stylized object images that emphasized shape representation. Together, by measuring the relative coding strength of object identity and nonidentity features, our approach provided a new tool to characterize feature coding in the human brain and the correspondence between the brain and CNNs. Copy rights belong to original authors. Visit the link for more info

Joint Representation of Color and Shape in Convolutional Neural Networks: A Stimulus-rich Network Perspective

rich cnn color shape representation copy joint stimulus xu biorxiv convolutional neural networks

Play Episode Listen Later Aug 12, 2020

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.08.11.246223v1?rss=1 Authors: Taylor, J., Xu, Y. Abstract: To interact with real-world objects, any effective visual system must jointly code the unique features defining each object. Despite decades of neuroscience research, we still lack a firm grasp on how the primate brain binds visual features. Here we apply a novel network-based stimulus-rich representational similarity approach to study color and shape binding in five convolutional neural networks (CNNs) with varying architecture, depth, and presence/absence of recurrent processing. All CNNs showed near-orthogonal color and shape processing in early layers, but increasingly interactive feature coding in higher layers, with this effect being much stronger for networks trained for object classification than untrained networks. These results characterize for the first time how multiple visual features are coded together in CNNs. The approach developed here can be easily implemented to characterize whether a similar coding scheme may serve as a viable solution to the binding problem in the primate brain. Copy rights belong to original authors. Visit the link for more info

The development of transformation tolerant visual representations differs between the human brain and convolutional neural networks

transformation development cnn visual copy tolerant human brain representations differs biorxiv convolutional neural networks

Play Episode Listen Later Aug 12, 2020

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.08.11.246934v1?rss=1 Authors: Xu, Y., Vaziri-Pashkam, M. Abstract: Existing single cell neural recording findings predict that, as information ascends the visual processing hierarchy in the primate brain, the relative similarity among the objects would be increasingly preserved across identity-preserving image transformations. Here we confirm this prediction and show that object category representational structure becomes increasingly invariant across position and size changes as information ascends the human ventral visual processing pathway. Such a representation, however, is not found in 14 different convolutional neural networks (CNNs) trained for object categorization that varied in architecture, depth and the presence/absence of recurrent processing. CNNs thus do not appear to form or maintain brain-like transformation-tolerant object identity representations at higher levels of visual processing despite the fact that CNNs may classify objects under various transformations. This limitation could potentially contribute to the large number of training data required to train CNNs and their limited ability to generalize to objects not included in training. Copy rights belong to original authors. Visit the link for more info

Explaining decisions of Graph Convolutional Neural Networks: patient-specific molecular subnetworks responsible for metastasis prediction in breast cancer

Play Episode Listen Later Aug 6, 2020

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.08.05.238519v1?rss=1 Authors: Chereda, H., Bleckmann, A., Menck, K., Perera-Bel, J., Stegmaier, P., Auer, F., Kramer, F., Leha, A., Beissbarth, T. Abstract: Motivation: Contemporary deep learning approaches show cutting-edge performance in a variety of complex prediction tasks. Nonetheless, the application of deep learning in healthcare remains limited since deep learning methods are often considered as non-interpretable black-box models. Layer-wise Relevance Propagation (LRP) is a technique to explain decisions of deep learning methods. It is widely used to interpret Convolutional Neural Networks (CNNs) applied on image data. Recently, CNNs started to extend towards non-euclidean domains like graphs. Molecular networks are commonly represented as graphs detailing interactions between molecules. Gene expression data can be assigned to the vertices of these graphs. In other words, gene expression data can be structured by utilizing molecular network information as prior knowledge. Graph-CNNs can be applied to structured gene expression data, for example, to predict metastatic events in breast cancer. Therefore, there is a need for explanations showing which part of a molecular network is relevant for predicting an event, e.g. distant metastasis in cancer, for each individual patient. Results: We extended the procedure of LRP to make it available for Graph-CNN and tested its applicability on a large breast cancer dataset. We present Graph Layer-wise Relevance Propagation (GLRP) as a new method to explain the decisions made by Graph-CNNs. We demonstrate a sanity check of the developed GLRP on a hand-written digits dataset, and then applied the method on gene expression data. We show that GLRP provides patient-specific molecular subnetworks that largely agree with clinical knowledge and identify common as well as novel, and potentially druggable, drivers of tumor progression. As a result this method could be potentially highly useful on interpreting classification results on the individual patient level, as for example in precision medicine approaches or a molecular tumor board. Copy rights belong to original authors. Visit the link for more info

predictions cnn patients decisions responsible explaining copy breast cancer kramer layer molecular graphs auer metastasis biorxiv lrp stegmaier convolutional neural networks

CoCoNet: Boosting RNA contact prediction by convolutional neural networks

predictions copy boosting ml rna dca pucci rnas biorxiv convolutional neural networks

Play Episode Listen Later Jul 31, 2020

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.07.30.229484v1?rss=1 Authors: Zerihun, M. B., Pucci, F., Schug, A. Abstract: Physics-based co-evolutionary models such as direct coupling analysis (DCA) in combination with machine learning (ML) techniques based on deep neural networks are able to predict protein contact maps with astonishing accuracy. Such contacts can be used as constraints in structure prediction and massively increase prediction accuracy. Unfortunately, the same ML methods cannot readily be applied to RNA as they rely on large structural datasets only available for proteins but not for RNAs. Here, we demonstrate how the small amount of data available for RNA can be used to significantly improve prediction of RNA contact maps. We introduce an algorithm called CoCoNet that is based on a combination of a Coevolutionary model and a shallow Convolutional Neural Network. Despite its simplicity and the small number of trained parameters, the method boosts the contact prediction accuracy by about 70% with respect to straightforward DCA as tested by cross-validation on a dataset of about sixty RNA structures. Both our extensive robustness tests and the limited number of parameters allow the generalization properties of our model. Finally, applications to other RNAs highlight the power of our approach. CoCoNet is freely available and can be found at https://github.com/KIT-MBS/coconet. Copy rights belong to original authors. Visit the link for more info

Semantic relatedness emerges in deep convolutional neural networks designed for object recognition

Play Episode Listen Later Jul 6, 2020

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.07.04.188169v1?rss=1 Authors: Huang, T., Zhen, Z., Liu, J. Abstract: Human not only can effortlessly recognize objects, but also characterize object categories into semantic concepts and construct nested hierarchical structures. Similarly, deep convolutional neural networks (DCNNs) can learn to recognize objects as perfectly as human; yet it is unclear whether they can learn semantic relatedness among objects that is not provided in the learning dataset. This is important because it may shed light on how human acquire semantic knowledge on objects without top-down conceptual guidance. To do this, we explored the relation among object categories, indexed by representational similarity, in two typical DCNNs (AlexNet and VGG11). We found that representations of object categories were organized in a hierarchical fashion, suggesting that the relatedness among objects emerged automatically when learning to recognize them. Critically, the emerged relatedness of objects in the DCNNs was highly similar to the WordNet in human, implying that top-down conceptual guidance may not be a prerequisite for human learning the relatedness among objects. Finally, the developmental trajectory of the relatedness among objects during training revealed that the hierarchical structure was constructed in a coarse-to-fine fashion, and evolved into maturity before the establishment of object recognition ability. Taken together, our study provides the first empirical evidence that semantic relatedness of objects emerged as a by-product of object recognition, implying that human may acquire semantic knowledge on objects without explicit top-down conceptual guidance. Copy rights belong to original authors. Visit the link for more info

deep designed copy liu critically semantic zhen relatedness biorxiv convolutional neural networks object recognition abstract human

Implementation-independent representation for deep convolutional neural networks and humans in processing faces

deep humans independent faces representation copy processing implementation liu xu biorxiv convolutional neural networks

Play Episode Listen Later Jun 29, 2020

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.06.26.171298v1?rss=1 Authors: Song, Y., Qu, Y., Xu, S., Liu, J. Abstract: Deep convolutional neural networks (DCNN) nowadays can match and even outperform human performance in challenging complex tasks. However, it remains unknown whether DCNNs achieve human-like performance through human-like processes; that is, do DCNNs use similar internal representations to achieve the task as humans? Here we applied a reverse-correlation method to reconstruct the internal representations when DCNNs and human observers classified genders of faces. We found that human observers and a DCNN pre-trained for face identification, VGG-Face, showed high similarity between their "classification images" in gender classification, suggesting similar critical information utilized in this task. Further analyses showed that the similarity of the representations was mainly observed at low spatial frequencies, which are critical for gender classification in human studies. Importantly, the prior task experience, which the VGG-Face was pre-trained for processing faces at the subordinate level (i.e., identification) as humans do, seemed necessary for such representational similarity, because AlexNet, a DCNN pre-trained to process objects at the basic level (i.e., categorization), succeeded in gender classification but relied on a completely different representation. In sum, although DCNNs and humans rely on different sets of hardware to process faces, they can use a similar representation, possibly from similar prior task experiences, to achieve the same computation goal. Therefore, our study provides the first empirical evidence supporting the hypothesis of implementation-independent representation. Copy rights belong to original authors. Visit the link for more info

Hierarchical sparse coding of objects in deep convolutional neural networks

deep copy coding objects liu hierarchical sparse zhen biorxiv convolutional neural networks

Play Episode Listen Later Jun 29, 2020

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.06.29.176032v1?rss=1 Authors: Liu, X., Zhen, Z., Liu, J. Abstract: Recently, deep convolutional neural networks (DCNNs) have attained human-level performances on challenging object recognition tasks owing to their complex internal representation. However, it remains unclear how objects are represented in DCNNs with an overwhelming number of features and non-linear operations. In parallel, the same question has been extensively studied in primates' brain, and three types of coding schemes have been found: one object is coded by entire neuronal population (distributed coding), or by one single neuron (local coding), or by a subset of neuronal population (sparse coding). Here we asked whether DCNNs adopted any of these coding schemes to represent objects. Specifically, we used the population sparseness index, which is widely-used in neurophysiological studies on primates' brain, to characterize the degree of sparseness at each layer in two representative DCNNs pretrained for object categorization, AlexNet and VGG11. We found that the sparse coding scheme was adopted at all layers of the DCNNs, and the degree of sparseness increased along the hierarchy. That is, the coding scheme shifted from distributed-like coding at lower layers to local-like coding at higher layers. Further, the degree of sparseness was positively correlated with DCNNs' performance in object categorization, suggesting that the coding scheme was related to behavioral performance. Finally, with the lesion approach, we demonstrated that both external learning experiences and built-in gating operations were necessary to construct such a hierarchical coding scheme. In sum, our study provides direct evidence that DCNNs adopted a hierarchically-evolved sparse coding scheme as the biological brain does, suggesting an implementation-independent principle of representing a myriad of objects efficiently. Copy rights belong to original authors. Visit the link for more info

Sebastian Ruder - Language independence and material properties

Underrated ML

Play Episode Listen Later Jun 15, 2020 94:05

This week we are joined by Sebastian Ruder. He is a research scientist at DeepMind, London. He has also worked at a variety of institutions such as AYLIEN, Microsoft, IBM's Extreme Blue, Google Summer of Code, and SAP. These experiences were completed in tangent with his studies which included studying Computational Linguistics at the University of Heidelberg, Germany and at Trinity College, Dublin before undertaking a PhD in Natural Language Processing and Deep Learning at the Insight Research Centre for Data Analytics. This week we discuss language independence and diversity in natural language processing whilst also taking a look at the attempts to identify material properties from images.As discussed in the podcast if you would like to donate to the current campaign of "CREATE DONATE EDUCATE" which supports Stop Hate UK then please find the link below: https://www.shorturl.at/glmszPlease also find additional links to help support black colleagues in the area of research;Black in AI twitter account: https://twitter.com/black_in_aiMentoring and proofreading sign-up to support our Black colleagues in research: https://twitter.com/le_roux_nicolas/status/1267896907621433344?s=20Underrated ML Twitter: https://twitter.com/underrated_mlSebastian Ruder Twitter: https://twitter.com/seb_ruderPlease let us know who you thought presented the most underrated paper in the form below: https://forms.gle/97MgHvTkXgdB41TC8Links to the papers:“On Achieving and Evaluating Language-Independence in NLP” - https://journals.linguisticsociety.org/elanguage/lilt/article/view/2624.html"The State and Fate of Linguistic Diversity and Inclusion in the NLP World” - https://arxiv.org/abs/2004.09095"Recognizing Material Properties from Images" - https://arxiv.org/pdf/1801.03127.pdfAdditional Links:Student perspectives on applying to NLP PhD programs: https://blog.nelsonliu.me/2019/10/24/student-perspectives-on-applying-to-nlp-phd-programs/Tim Dettmer's post on how to pick your grad school: https://timdettmers.com/2020/03/10/how-to-pick-your-grad-school/Rachel Thomas' blog post on why you should blog: https://medium.com/@racheltho/why-you-yes-you-should-blog-7d2544ac1045Emily Bender's The Gradient article: https://thegradient.pub/the-benderrule-on-naming-the-languages-we-study-and-why-it-matters/Paper on order-sensitive vs order-free methods: https://www.aclweb.org/anthology/N19-1253.pdf"Exploring the Origins and Prevalence of Texture Bias in Convolutional Neural Networks": https://arxiv.org/abs/1911.09071Sebastian's website where you can find all his blog posts: https://ruder.io/

Convolutional Neural Networks

Linear Digressions

Play Episode Listen Later May 31, 2020 21:55

This is a re-release of an episode that originally aired on April 1, 2018 If you've done image recognition or computer vision tasks with a neural network, you've probably used a convolutional neural net. This episode is all about the architecture and implementation details of convolutional networks, and the tricks that make them so good at image tasks.

learning science data linear digressions convolutional neural networks

Predicting Cortical Signatures of Consciousness using Dynamic Functional Connectivity Graph-Convolutional Neural Networks

cnn consciousness dynamic copy predicting functional connectivity graphs signatures cortical biorxiv uhrig convolutional neural networks

Play Episode Listen Later May 12, 2020

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.05.11.078535v1?rss=1 Authors: Grigis, A., Tasserie, J., Frouin, V., Jarraya, B., Uhrig, L. Abstract: Decoding the levels of consciousness from cortical activity recording is a major challenge in neuroscience. Using clustering algorithms, we previously demonstrated that resting-state functional MRI (rsfMRI) data can be split into several clusters also called "brain states" corresponding to "functional configurations" of the brain. Here, we propose to use a supervised machine learning method based on artificial neural networks to predict functional brain states across levels of consciousness from rsfMRI. Because it is key to consider the topology of brain regions used to build the dynamical functional connectivity matrices describing the brain state at a given time, we applied BrainNetCNN, a graph-convolutional neural network (CNN), to predict the brain states in awake and anesthetized non-human primate rsfMRI data. BrainNetCNN achieved a high prediction accuracy that lies in [0.674, 0.765] depending on the experimental settings. We propose to derive the set of connections found to be important for predicting a brain state, reflecting the level of consciousness. The results demonstrate that deep learning methods can be used not only to predict brain states but also to provide additional insight on cortical signatures of consciousness with potential clinical consequences for the monitoring of anesthesia and the diagnosis of disorders of consciousness. Copy rights belong to original authors. Visit the link for more info

Convolutional Neural Networks for IoT Time Series Data

The Private Equity Digital Transformation Show

Play Episode Listen Later Dec 6, 2019 48:13

Convoluted Neural Networks or CNNs are a type of AI typically used in computer vision to process images, but they are also applicable to process the time series data we typically get from sensors in IoT. In this episode of the IoT show, I speak with Simon Crosby about how these CNNs can be used to make predictions about the future and reduce the massive amounts of data we collect to just the important stuff. Read the rest of the show analysis notes including the transcripts at: http://bit.ly/IoTPodcast109notes This show is brought to you by DIGITAL OPERATING PARTNERS Related links you may find useful: Season 1: Episodes and show notes Season 1 book: IoT Inc Season 2: Episodes and show notes Season 2 book: The Private Equity Digital Operating Partner Training: Digital transformation certification

learning ai deep data tools cnn iot predictive time series convolutional neural networks

58. (Tech Talk) How to Apply Deep Learning Models to Image Classification

Data Couture

Play Episode Listen Later Sep 25, 2019 21:21

On this first episode of Tech Talk, we cover one of the classics of deep learning and artificial neural networks: Image Classification!Check out this walkthrough of how to apply Convolutional Neural Networks to a real-life feeding problem!To keep up with the podcast be sure to visit our website at datacouture.org, follow us on twitter @datacouturepod, and on instagram @datacouturepodcast. And, if you'd like to help support future episodes, then consider becoming a patron at patreon.com/datacouture!Support the show (https://www.patreon.com/datacouture)

ai cnn models ml classification tech talk deep learning predictive analytics convolutional neural networks

Yann LeCun: Deep Learning, Convolutional Neural Networks, and Self-Supervised Learning

Lex Fridman Podcast

Play Episode Listen Later Aug 31, 2019 76:07

Yann LeCun is one of the fathers of deep learning, the recent revolution in AI that has captivated the world with the possibility of what machines can learn from data. He is a professor at New York University, a Vice President & Chief AI Scientist at Facebook, co-recipient of the Turing Award for his work on deep learning. He is probably best known as the founder of convolutional neural networks, in particular their early application to optical character recognition. This conversation is part of the Artificial Intelligence podcast. If you would like to get more information about this podcast go to

ai vice president artificial intelligence new york university deep learning yann lecun turing award supervised learning convolutional neural networks

Deep Learning for Population Genetic Inference with Dan Schrider - TWiML Talk #249

This Week in Machine Learning & Artificial Intelligence (AI) Podcast

Play Episode Listen Later Apr 8, 2019 49:53

Today we’re joined by Dan Schrider, assistant professor in the department of genetics at The University of North Carolina at Chapel Hill. My discussion with Dan starts with an overview of population genomics and from there digs into his application of machine learning in the field, allowing us to, for example, better understand population size changes and gene flow from DNA sequences. We then dig into Dan’s paper “The Unreasonable Effectiveness of Convolutional Neural Networks in Population Genetic Inference,” which was published in the Molecular Biology and Evolution journal, which examines the idea that CNNs are capable of outperforming expert-derived statistical methods for some key problems in the field. Thanks to Pegasystems for sponsoring today's show! I'd like to invite you to join me at PegaWorld, the company’s annual digital transformation conference, which takes place this June in Las Vegas. To learn more about the conference or to register, visit pegaworld.com and use TWIML19 in the promo code field when you get there for $200 off. The complete show notes for this episode can be found at https://twimlai.com/talk/249.

Building a Visual Search Engine for Art Historians with Benoit Seguin, Computer Scientist

State Of The Art

Play Episode Listen Later Oct 4, 2018 48:31

In an age of rapidly evolving tech facilitating a number of things from daily tasks, to communication and research, some subjects trail behind, notably, art history. Replica aims to facilitate art historical research by using machine learning and modern image searching algorithms to help art historians navigate large iconographic collections. In short, Replica aims to go beyond the limitations of search terms and keywords and use images to enable researchers to identify visual information which may not been indexed--textual queries cannot, for example, show results for similar shapes, forms, or motifs. Replica could assist researchers in identifying visual links, pin-pointing when a specific type of iconography emerged and how it has evolved over time.In this episode, we speak with Benoit Seguin, a computer scientist helping build the Replica search engine. Seguin explains how Replica works, what the mission is, who's involved, and how it can be immensely useful to those interested in art, art history, and visual culture.-About Replica-The Replica project led by the DHLAB aims to build the first search engine designed specifically for the search and exploration of artistic collections (including paintings, drawings, engravings, sculpture and photography). This employs the latest state-of-the art artificial intelligence techniques, such as Deep Learning and Convolutional Neural Networks, for the search and display of information. In partnership with the Giorgio Cini Foundation in Venice and Factum Arte in Madrid, the Replica project aims to digitize roughly one million artistic reproductions using these images to populate the new search engine and as the basis for new art historical inquiries. -About Benoit Seguin- Benoit Seguin is a PhD student at the Digital Humanities Laboratory (DHLAB) at EPFL. His main interests lie in Computer Vision, Machine Learning and Image Processing. Benoit's thesis is based on the Replica Project where he implements machine learning and modern image searching algorithms to help art historians navigate large, iconographic collections. Benoit received a Master of Science from EPFL and a Diplôme d’Ingénieur from École Polytechnique.

MCL305: Scaling Convolutional Neural Networks with Kubernetes and TensorFlow on AWS

AWS re:Invent 2017

Play Episode Listen Later Nov 30, 2017 36:55

In this session, Reza Zadeh, CEO of Matroid, presents a Kubernetes deployment on Amazon Web Services that provides customized computer vision to a large number of users. Reza offers an overview of Matroid's pipeline and demonstrates how to customize computer vision neural network models in the browser, followed by building, training, and visualizing TensorFlow models, which are provided at scale to monitor video streams.

ceo scaling amazon web services reza kubernetes tensorflow convolutional neural networks reza zadeh

025 Convolutional Neural Networks

Machine Learning Guide

Play Episode Listen Later Oct 30, 2017 44:21

Convnets or CNNs. Filters, feature maps, window/stride/padding, max-pooling. ocdevel.com/mlg/25 for notes and resources

learning ai cnn intelligence artificial ml filters convolutional neural networks

Me and My AI 7: Convolutional Neural Networks

Me and My AI

Play Episode Listen Later Sep 2, 2017 21:03

Convolutional Neural Networks: this is CNN (but not the news network). A look at the AI technology behind a recent explosion in the effectiveness of artificial image recognition. SHOW NOTES 1. Neural Networks Made Easy

ai cnn convolutional neural networks

Cardiologist Level Arrhythmia Detection with CNNs

Data Skeptic

Play Episode Listen Later Aug 25, 2017 32:05

Our guest Pranav Rajpurkar and his coauthored recently published Cardiologist-Level Arrhythmia Detection with Convolutional Neural Networks, a paper in which they demonstrate the use of Convolutional Neural Networks which outperform board certified cardiologists in detecting a wide range of heart arrhythmias from ECG data.

detection cardiologists ecg arrhythmia convolutional neural networks

Episode 008 - Lesson 4 - Part 1 (Practical Deep Learning for Coders)

Startup Data Science

Play Episode Listen Later Jul 8, 2017 31:26

Apurva loved Jeremy's presentation using Excel to show how calculations are being made; it was a great confidence-building exercise for her to replicate it in Excel. Edderic's excited about Jeremy's claim that Convolutional Neural Networks are doing well in Speech Recognition. There are tons of machine learning algorithms out there; he thinks it would be nice to have just one super algorithm/architecture to rule them all. Alex explains his idea of convolution through an analogy.

lesson excel coders speech recognition apurva convolutional neural networks practical deep learning

07 - Capturing Semantic Similarity for Entity Linking with Convolutional Neural Networks

NLP Highlights

Play Episode Listen Later May 23, 2017 14:27

https://www.semanticscholar.org/paper/Capturing-Semantic-Similarity-for-Entity-Linking-w-Francis-Landau-Durrett/1c9aca60f7ac5edcceb73d612806704a7d662643

capturing linking entity semantic similarity convolutional neural networks

[MINI] Convolutional Neural Networks

Data Skeptic

Play Episode Listen Later May 19, 2017 14:54

CNNs are characterized by their use of a group of neurons typically referred to as a filter or kernel. In image recognition, this kernel is repeated over the entire image. In this way, CNNs may achieve the property of translational invariance - once trained to recognize certain things, changing the position of that thing in an image should not disrupt the CNN's ability to recognize it. In this episode, we discuss a few high-level details of this important architecture.

cnn convolutional neural networks

Energy-Efficient Hardware for Embedded Vision and Deep Convolutional Neural Networks 5/1/2017

Computer Architecture Seminar Series

Play Episode Listen Later May 12, 2017 61:32

Visual object detection and recognition are needed for a wide range of applications including robotics/drones, self-driving cars, smart Internet of Things, and portable/wearable electronics. For many of these applications, local embedded processing is preferred due to privacy or latency concerns. In this talk, we will describe how joint algorithm and hardware design can be used to reduce the energy consumption of object detection and recognition while delivering real-time and robust performance. We will discuss several energy-efficient techniques that exploit sparsity, reduce data movement and storage costs, and show how they can be applied to popular forms of object detection and recognition, including those that use deep convolutional neural nets. We will present results from recently fabricated ASICs (e.g. our deep CNN accelerator named “Eyeriss”) that demonstrate these techniques in real-time computer vision systems. Speaker Biography Vivienne Sze is an Assistant Professor at MIT in the Electrical Engineering and Computer Science Department. Her research interests include energy-aware signal processing algorithms, and low-power circuit and system design for multimedia applications. Prior to joining MIT, she was a Member of Technical Staff in the R&D Center at TI, where she developed algorithms and hardware for the latest video coding standard H.265/HEVC. She is a co-editor of the book entitled “High Efficiency Video Coding (HEVC): Algorithms and Architectures” (Springer, 2014). Dr. Sze received the B.A.Sc. degree from the University of Toronto in 2004, and the S.M. and Ph.D. degree from MIT in 2006 and 2010, respectively. In 2011, she was awarded the Jin-Au Kong Outstanding Doctoral Thesis Prize in electrical engineering at MIT for her thesis on “Parallel Algorithms and Architectures for Low Power Video Decoding”. She is a recipient of the 2016 AFOSR Young Investigator Award, the 2016 3M Non-tenured Faculty Award, the 2014 DARPA Young Faculty Award, the 2007 DAC/ISSCC Student Design Contest Award and a co-recipient of the 2008 A-SSCC Outstanding Design Award. For more information about research in the Energy-Efficient Multimedia Systems Group at MIT visit: http://www.rle.mit.edu/eems/

Convolutional Neural Networks with Matt Zeiler

Machine Learning – Software Engineering Daily

Play Episode Listen Later May 10, 2017 54:37

Convolutional neural networks are a machine learning tool that uses layers of convolution and pooling to process and classify inputs. CNNs are useful for identifying objects in images and video. In this episode, we focus on the application of convolutional neural networks to image and video recognition and classification. Matt Zeiler is the CEO of The post Convolutional Neural Networks with Matt Zeiler appeared first on Software Engineering Daily.

cnn zeiler software engineering daily convolutional convolutional neural networks

Virtual EEG-electrodes: Convolutional neural networks as a method for upsampling or restoring channels