Unsupervised neural network models of the ventral visual stream

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

---

The neuroconnectionist research programme

Nature reviews. Neuroscience, Journal Year: 2023, Volume and Issue: 24(7), P. 431 - 450

Published: May 30, 2023

Language: Английский

---

An ecologically motivated image dataset for deep learning yields better models of human vision

Proceedings of the National Academy of Sciences, Journal Year: 2021, Volume and Issue: 118(8)

Published: Feb. 15, 2021

Significance Inspired by core principles of information processing in the brain, deep neural networks (DNNs) have demonstrated remarkable success computer vision applications. At same time, trained on task object classification exhibit similarities to representations found primate visual system. This result is surprising because datasets commonly used for training are designed be engineering challenges. Here, we use linguistic corpus statistics and human concreteness ratings as guiding design a resource that more closely mirrors categories relevant humans. The ecoset, collection 1.5 million images from 565 basic-level categories. We show ecoset-trained DNNs yield better models higher-level cortex behavior.

Language: Английский

---

A self-supervised domain-general learning framework for human ventral stream representation

Nature Communications, Journal Year: 2022, Volume and Issue: 13(1)

Published: Jan. 25, 2022

Abstract Anterior regions of the ventral visual stream encode substantial information about object categories. Are top-down category-level forces critical for arriving at this representation, or can representation be formed purely through domain-general learning natural image structure? Here we present a fully self-supervised model which learns to represent individual images, rather than categories, such that views same are embedded nearby in low-dimensional feature space, distinctly from other recently encountered views. We find category implicitly emerges local similarity structure space. Further, these models learn hierarchical features capture brain responses across human stream, on par with category-supervised models. These results provide computational support framework guiding formation where proximate goal is not explicitly information, but instead unique, compressed descriptions world.

Language: Английский

---

Grounded language acquisition through the eyes and ears of a single child

Science, Journal Year: 2024, Volume and Issue: 383(6682), P. 504 - 511

Published: Feb. 1, 2024

Starting around 6 to 9 months of age, children begin acquiring their first words, linking spoken words visual counterparts. How much this knowledge is learnable from sensory input with relatively generic learning mechanisms, and how requires stronger inductive biases? Using longitudinal head-mounted camera recordings one child aged 25 months, we trained a neural network on 61 hours correlated visual-linguistic data streams, feature-based representations cross-modal associations. Our model acquires many word-referent mappings present in the child’s everyday experience, enables zero-shot generalization new referents, aligns its linguistic conceptual systems. These results show critical aspects grounded word meaning are through joint representation associative input.

Language: Английский

---

Intuitive physics learning in a deep-learning model inspired by developmental psychology

Nature Human Behaviour, Journal Year: 2022, Volume and Issue: 6(9), P. 1257 - 1267

Published: July 11, 2022

'Intuitive physics' enables our pragmatic engagement with the physical world and forms a key component of 'common sense' aspects thought. Current artificial intelligence systems pale in their understanding intuitive physics, comparison to even very young children. Here we address this gap between humans machines by drawing on field developmental psychology. First, introduce open-source machine-learning dataset designed evaluate conceptual adopting violation-of-expectation (VoE) paradigm from Second, build deep-learning system that learns physics directly visual data, inspired studies cognition We demonstrate model can learn diverse set concepts, which depends critically object-level representations, consistent findings consider implications these results both for AI research human cognition.

Language: Английский

---

Measuring and modeling the motor system with machine learning

Current Opinion in Neurobiology, Journal Year: 2021, Volume and Issue: 70, P. 11 - 23

Published: June 8, 2021

The utility of machine learning in understanding the motor system is promising a revolution how to collect, measure, and analyze data. field movement science already elegantly incorporates theory engineering principles guide experimental work, this review we discuss growing use learning: from pose estimation, kinematic analyses, dimensionality reduction, closed-loop feedback, its neural correlates untangling sensorimotor systems. We also give our perspective on new avenues where markerless motion capture combined with biomechanical modeling networks could be platform for hypothesis-driven research.

Language: Английский

---

The developmental trajectory of object recognition robustness: Children are like small adults but unlike big deep neural networks

Journal of Vision, Journal Year: 2023, Volume and Issue: 23(7), P. 4 - 4

Published: July 6, 2023

In laboratory object recognition tasks based on undistorted photographs, both adult humans and deep neural networks (DNNs) perform close to ceiling. Unlike adults’, whose performance is robust against a wide range of image distortions, DNNs trained standard ImageNet (1.3M images) poorly distorted images. However, the last 2 years have seen impressive gains in DNN distortion robustness, predominantly achieved through ever-increasing large-scale datasets—orders magnitude larger than ImageNet. Although this simple brute-force approach very effective achieving human-level robustness DNNs, it raises question whether human too, simply due extensive experience with (distorted) visual input during childhood beyond. Here we investigate by comparing core 146 children (aged 4–15 years) adults DNNs. We find, first, that already 4- 6-year-olds show remarkable distortions outperform Second, estimated number images had been exposed their lifetime. Compared various children’s high requires relatively little data. Third, when recognizing objects, children—like but unlike DNNs—rely heavily shape not texture cues. Together our results suggest emerges early developmental trajectory unlikely result mere accumulation input. Even though current match regarding they seem rely different more data-hungry strategies do so.

Language: Английский

---

Learning high-level visual representations from a child’s perspective without strong inductive biases

Nature Machine Intelligence, Journal Year: 2024, Volume and Issue: 6(3), P. 271 - 283

Published: March 7, 2024

Language: Английский

---

Contrastive learning explains the emergence and function of visual category-selective regions

Science Advances, Journal Year: 2024, Volume and Issue: 10(39)

Published: Sept. 25, 2024

Modular and distributed coding theories of category selectivity along the human ventral visual stream have long existed in tension. Here, we present a reconciling framework—contrastive coding—based on series analyses relating within biological artificial neural networks. We discover that, models trained with contrastive self-supervised objectives over rich natural image diet, category-selective tuning naturally emerges for faces, bodies, scenes, words. Further, lesions these model units lead to selective, dissociable recognition deficits, highlighting their distinct functional roles information processing. Finally, pre-identified can predict responses all corresponding face-, scene-, body-, word-selective regions cortex, under highly constrained sparse positive encoding procedure. The success this single indicates that brain-like specialization emerge without category-specific learning pressures, as system learns untangle content. Contrastive coding, therefore, provides unifying account object emergence representation brain.

Language: Английский

---