A deep learning framework for neuroscience

By Blake Aaron Richards, Timothy P. Lillicrap, Philippe Beaudoin, Yoshua Bengio, Rafal Bogacz, A. Christensen, Claudia Clopath, Rui Ponte Costa, Archy O. de Berker, Surya Ganguli and 22 more. 2019

1,016 citations recorded by the source corpus.

Systems neuroscience seeks explanations for how the brain implements a wide variety of perceptual, cognitive and motor tasks. Conversely, artificial intelligence attempts to design computational systems based on the tasks they will have to solve. In artificial neural networks, the three components specified by design are the objective functions, the learning rules and the architectures. With the growing success of deep learning, which utilizes brain-inspired architectures, these three designed components have increasingly become central to how we model, engineer and optimize complex artificial learning systems. Here we argue that a greater focus on these components would also benefit systems neuroscience. We give examples of how this optimization-based framework can drive theoretical and experimental progress in neuroscience. We contend that this principled perspective on systems neuroscience will help to generate more rapid progress. A deep network is best understood in terms of components used to design it—objective functions, architecture and learning rules—rather than unit-by-unit computation. Richards et al. argue that this inspires fruitful approaches to systems neuroscience.

Subfields: Cognitive Neuroscience, Electrical and Electronic Engineering

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Cognitive Neuroscience

In Electrical and Electronic Engineering

Citations in the corpus

Orthogonal representations for robust context-dependent task performance in brains and neural networks (2022, positive): “There has been a recent resurgence of interest in neural networks (or ‘‘deep learning models’’) as computational theories of biological brains (Richards et al., 2019; Saxe et al., 2021).”
Learning from the Past: Meta-Continual Learning with Knowledge Embedding for Jointly Sketch, Cartoon, and Caricature Face Recognition (2020, background): “In particular, they learn referents in knowledge by statistically matching words with occurrences of images in the environment [40].”
Direct-fit to nature: an evolutionary perspective on biological (and artificial) neural networks (2019, methodology): “Across species and models, BNNs and ANNs can differ considerably in their circuit architecture, learning rules, and objective functions ( Richards et al., 2019).”
Training Spiking Neural Networks Using Lessons From Deep Learning (2021, background): “While it is unlikely that our brains use something as explicit as a cross-entropy loss function, it is fair to say that humans and animals have baseline objectives [121].”
Convolutional Neural Networks as a Model of the Visual System: Past, Present, and Future (2020, methodology): “For this reason, an alternative way to describe the network that is compact without relying on simple statements about computations has been proposed (Lillicrap & Kording, 2019; Richards et al., 2019); this viewpoint focuses on describing the architec-”
Unsupervised neural network models of the ventral visual stream (2020, background): “How can these gaps be bridged? Deep learning systems in neuroscience can be thought of as having several basic components (22, 72): an architecture class capturing neuroanatomical knowledge; an objective function capturing hypotheses about the signals driving learning; a training dataset capturing t”
Shared computational principles for language processing in humans and deep language models (2022, background): “To the extent that the brain relies on a next-word prediction objective to learn how to use language in context, it likely uses a different implementation 54 .”
Artificial Neural Networks for Neuroscientists: A Primer (2020, background): “, 2014), and insights from the perspective of biological evolution (Lindsey et al., 2019; Richards et al., 2019).”
If deep learning is the answer, what is the question? (2020, background): “The claim has also been made that modelling brains as neural networks relieves researchers of the burden of exhaustively documenting and interpreting the coding properties of single neurons 33 .”
A connectome of the Drosophila central complex reveals network motifs suitable for flexible navigation and context-dependent action selection (2020, background): “…learning, a broader question concerns the level of understanding that is appropriate or even possible for the function of large and complex neural networks (Gao and Ganguli, 2015; Hasson et al., 2020; Lillicrap and Kording, 2019; Richards et al., 2019; Saxe et al., 2020; Yamins and DiCarlo, 2016).”