Unsupervised deep learning identifies semantic disentanglement in single inferotemporal face patch neurons DOI Creative Commons
Irina Higgins, Le Chang,

Victoria Langston

et al.

Nature Communications, Journal Year: 2021, Volume and Issue: 12(1)

Published: Nov. 9, 2021

Abstract In order to better understand how the brain perceives faces, it is important know what objective drives learning in ventral visual stream. To answer this question, we model neural responses faces macaque inferotemporal (IT) cortex with a deep self-supervised generative model, β -VAE, which disentangles sensory data into interpretable latent factors, such as gender or age. Our results demonstrate strong correspondence between factors discovered by -VAE and those coded single IT neurons, beyond that found for baselines, including handcrafted state-of-the-art of face perception, Active Appearance Model, classifiers. Moreover, able reconstruct novel images using signals from just handful cells. Together our imply optimising disentangling leads representations closely resemble at unit level. This points plausible brain.

Language: Английский

2022 roadmap on neuromorphic computing and engineering DOI Creative Commons
Dennis Valbjørn Christensen, Regina Dittmann, B. Linares-Barranco

et al.

Neuromorphic Computing and Engineering, Journal Year: 2022, Volume and Issue: 2(2), P. 022501 - 022501

Published: Jan. 12, 2022

Abstract Modern computation based on von Neumann architecture is now a mature cutting-edge science. In the architecture, processing and memory units are implemented as separate blocks interchanging data intensively continuously. This transfer responsible for large part of power consumption. The next generation computer technology expected to solve problems at exascale with 10 18 calculations each second. Even though these future computers will be incredibly powerful, if they type architectures, consume between 20 30 megawatts not have intrinsic physically built-in capabilities learn or deal complex our brain does. These needs can addressed by neuromorphic computing systems which inspired biological concepts human brain. new has potential used storage amounts digital information much lower consumption than conventional processors. Among their applications, an important niche moving control from centers edge devices. aim this roadmap present snapshot state provide opinion challenges opportunities that holds in major areas technology, namely materials, devices, circuits, algorithms, ethics. collection perspectives where leading researchers community own view about current research area. We hope useful resource providing concise yet comprehensive introduction readers outside field, those who just entering well established community.

Language: Английский

Citations

462
Convolutional Neural Networks as a Model of the Visual System: Past, Present, and Future DOI
Grace W. Lindsay

Journal of Cognitive Neuroscience, Journal Year: 2020, Volume and Issue: 33(10), P. 2017 - 2031

Published: Feb. 6, 2020

Abstract Convolutional neural networks (CNNs) were inspired by early findings in the study of biological vision. They have since become successful tools computer vision and state-of-the-art models both activity behavior on visual tasks. This review highlights what, context CNNs, it means to be a good model computational neuroscience various ways can provide insight. Specifically, covers origins CNNs methods which we validate them as It then goes elaborate what learn about understanding experimenting discusses emerging opportunities for use research beyond basic object recognition.

Language: Английский

Citations

442
If deep learning is the answer, what is the question? DOI
Andrew Saxe, Stephanie Nelli, Christopher Summerfield

et al.

Nature reviews. Neuroscience, Journal Year: 2020, Volume and Issue: 22(1), P. 55 - 67

Published: Nov. 16, 2020

Language: Английский

Citations

342
Direct Fit to Nature: An Evolutionary Perspective on Biological and Artificial Neural Networks DOI Creative Commons
Uri Hasson, Samuel A. Nastase, Ariel Goldstein

et al.

Neuron, Journal Year: 2020, Volume and Issue: 105(3), P. 416 - 434

Published: Feb. 1, 2020

Language: Английский

Citations

304
A connectome of the Drosophila central complex reveals network motifs suitable for flexible navigation and context-dependent action selection DOI Creative Commons
Brad K. Hulse, Hannah Haberkern, Romain Franconville

et al.

eLife, Journal Year: 2021, Volume and Issue: 10

Published: Oct. 26, 2021

Flexible behaviors over long timescales are thought to engage recurrent neural networks in deep brain regions, which experimentally challenging study. In insects, circuit dynamics a region called the central complex (CX) enable directed locomotion, sleep, and context- experience-dependent spatial navigation. We describe first complete electron microscopy-based connectome of

Language: Английский

Citations

304
Shared computational principles for language processing in humans and deep language models DOI Creative Commons
Ariel Goldstein, Zaid Zada,

Eliav Buchnik

et al.

Nature Neuroscience, Journal Year: 2022, Volume and Issue: 25(3), P. 369 - 380

Published: March 1, 2022

Departing from traditional linguistic models, advances in deep learning have resulted a new type of predictive (autoregressive) language models (DLMs). Using self-supervised next-word prediction task, these generate appropriate responses given context. In the current study, nine participants listened to 30-min podcast while their brain were recorded using electrocorticography (ECoG). We provide empirical evidence that human and autoregressive DLMs share three fundamental computational principles as they process same natural narrative: (1) both are engaged continuous before word onset; (2) match pre-onset predictions incoming calculate post-onset surprise; (3) rely on contextual embeddings represent words contexts. Together, our findings suggest biologically feasible framework for studying neural basis language.

Language: Английский

Citations

289
Artificial Neural Networks for Neuroscientists: A Primer DOI Creative Commons
Guangyu Robert Yang, Xiao‐Jing Wang

Neuron, Journal Year: 2020, Volume and Issue: 107(6), P. 1048 - 1070

Published: Sept. 1, 2020

Language: Английский

Citations

285
Unsupervised neural network models of the ventral visual stream DOI Creative Commons
Chengxu Zhuang, Siming Yan, Aran Nayebi

et al.

Proceedings of the National Academy of Sciences, Journal Year: 2021, Volume and Issue: 118(3)

Published: Jan. 11, 2021

Significance Primates show remarkable ability to recognize objects. This is achieved by their ventral visual stream, multiple hierarchically interconnected brain areas. The best quantitative models of these areas are deep neural networks trained with human annotations. However, they receive more annotations than infants, making them implausible the stream development. Here, we report that recent progress in unsupervised learning has largely closed this gap. We find learned methods achieve prediction accuracy equals or exceeds today’s models. These results illustrate a use model system and present strong candidate for biologically plausible computational theory sensory learning.

Language: Английский

Citations

263
Keep it real: rethinking the primacy of experimental control in cognitive neuroscience DOI Creative Commons
Samuel A. Nastase, Ariel Goldstein, Uri Hasson

et al.

NeuroImage, Journal Year: 2020, Volume and Issue: 222, P. 117254 - 117254

Published: Aug. 13, 2020

Naturalistic experimental paradigms in neuroimaging arose from a pressure to test the validity of models we derive highly-controlled experiments real-world contexts. In many cases, however, such efforts led realization that developed under particular manipulations failed capture much variance outside context manipulation. The critique non-naturalistic is not recent development; it echoes persistent and subversive thread history modern psychology. brain has evolved guide behavior multidimensional world with interacting variables. assumption artificially decoupling manipulating these variables will lead satisfactory understanding may be untenable. We develop an argument for primacy naturalistic paradigms, point developments machine learning as example transformative power relinquishing control. should deployed afterthought if hope build extend beyond laboratory into real world.

Language: Английский

Citations

257
Illuminating dendritic function with computational models DOI
Panayiota Poirazi, Athanasia Papoutsi

Nature reviews. Neuroscience, Journal Year: 2020, Volume and Issue: 21(6), P. 303 - 321

Published: May 11, 2020

Language: Английский

Citations

245