Context-dependent computation by recurrent dynamics in prefrontal cortex
By Valerio Mante, David Sussillo, Krishna V. Shenoy, William T. Newsome. 2013
1,875 citations recorded by the source corpus.
Prefrontal cortex is thought to have a fundamental role in flexible, context-dependent behaviour, but the exact nature of the computations underlying this role remains largely unknown. In particular, individual prefrontal neurons often generate remarkably complex responses that defy deep understanding of their contribution to behaviour. Here we study prefrontal cortex activity in macaque monkeys trained to flexibly select and integrate noisy sensory inputs towards a choice. We find that the observed complexity and functional roles of single neurons are readily understood in the framework of a dynamical process unfolding at the level of the population. The population dynamics can be reproduced by a trained recurrent neural network, which suggests a previously unknown mechanism for selection and integration of task-relevant inputs. This mechanism indicates that selection and integration are two aspects of a single dynamical process unfolding within the same prefrontal circuits, and potentially provides a novel, general framework for understanding context-dependent computations.
Subfields: Cognitive Neuroscience, Neuroscience
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
- Xiao-Jing Wang
- Guangyu Robert Yang
- Rafael Yuste
- Srdjan Ostojic
- Mehrdad Jazayeri
- John P. Cunningham
- Matthew Botvinick
- Friedemann Zenke
In Neuroscience
- Rafael Yuste
- Guangyu Robert Yang
- Xiao-Jing Wang
- Rahaf Aljundi
- John P. Cunningham
- Gido M. van de Ven
- Tinne Tuytelaars
- Bahram Zonooz
Citations in the corpus
- Overcoming catastrophic forgetting in neural networks (2016, background): “…as supporting these capabilities by sustaining neural representations of task context that exert top-down gating influences on sensory processing, working memory, and action selection in lower-level regions [O’Reilly and Frank, 2006, Mante et al., 2013, Miller and Cohen, 2001, Doya et al., 2002].”
- Task-Aware Information Routing from Common Representation Space in Lifelong Learning (2023, background): “The prefrontal cortex of the primate brain is presumed to have task-dependent neural representations that act as a gating in different brain functions (Mante et al., 2013).”
- Orthogonal representations for robust context-dependent task performance in brains and neural networks (2022, positive): “We assessed the embedding dimensionality of the patterns observed in monkey FEF using the same truncated SVD approach described above for the human fMRI data.”
- Natural continual learning: success is a journey, not (just) a destination (2021, methodology): “To do this, we analyzed latent representations of the RNN activity trajectories, as is commonly done to study the collective dynamics of artificial and biological networks [15, 18, 19, 28, 51].”
- A modeling framework for adaptive lifelong learning with transfer and savings through gating in the prefrontal cortex (2020, background): “In our model, each “unit” within our recurrent networks represents a population of neurons, and, as such, inputs to the network were represented by populations of neurons, each of which receives multiple inputs and responds with mixed selectivity as has been observed in PFC (57, 58).”
- Continual learning with hypernetworks (2019, neutral): “This interpretation is consistent with a growing body of work which suggests the involvement of modulatory inputs in implementing context- or task-dependent network mode-switching [29, 30, 31].”
- Expert Gate: Lifelong Learning with a Network of Experts (2016, background): “In fact, also the prefrontal cortex of the primate brain is considered to have neural representations of task context that act as a gating in different brain functions [23].”
- Capacity and Trainability in Recurrent Neural Networks (2016, background): “…application domains in deep learning (speech recognition (Amodei et al., 2015), seq2seq (Sutskever et al., 2014), neural machine translation (Bahdanau et al., 2014), the DRAW model (Gregor et al., 2015), educational applications (Piech et al., 2015), and scientific discovery (Mante et al., 2013)).”
- On the importance of single directions for generalization (2018, background): “ stimuli and have shown that task information can be decoded with high accuracy from populations of these neurons with low individual class selectivity (Averbeck et al. (2006), Rigotti et al. (2013), Mante et al. (2013), Raposo et al. (2014), Morcos & Harvey (2016), Zylberberg (2017)). Perturbat”
- Continuous Learning of Context-dependent Processing in Neural Networks (2018, background): “In addition to the role of supporting fast concept learning, another function of the PFC is to represent the contextual information in the form of working memory [10], which guides the selection of the sensory features that are most relevant for the current task [6].”
- A dataset and architecture for visual reasoning with a working memory (2018, background): “One major goal of this dataset is to build a compositional set of tasks that include variants of many cognitive tasks studied in humans and other animals [32,26,33,34,28,29,30,31] (see also Introduction and Related Work).”
- Flexible sensorimotor computations through rapid reconfiguration of cortical dynamics (2018, positive): “Similarly, the presence of a context input can enable a recurrent neural network to perform flexible rule-(Mante et al. 2013; Song et al. 2016) and category-based decisions (Chaisangmongkon et al. 2017).”
- Linking Connectivity, Dynamics, and Computations in Low-Rank Recurrent Neural Networks. (2017, neutral): “We focus on four tasks of increasing complexity, starting with basic binary discrimination and ending with context-dependent evidence integration (Mante et al., 2013).”
- Motor cortex embeds muscle-like commands in an untangled population response (2018, neutral): “…body of work seeks to understand neural responses in terms of signals that a recurrent or feedback-driven neural network would need to perform the relevant task (Hennequin et al., 2014; Li et al., 2016; Lillicrap and Scott, 2013; Mante et al., 2013; Michaels et al., 2016; Sussillo and Barak, 2013).”