Why neurons mix: high dimensionality for higher cognition.

By Stefano Fusi, Earl K. Miller, Mattia Rigotti. 2016

750 citations recorded by the source corpus.

Subfields: Artificial Intelligence, 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 Artificial Intelligence

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): “Previous analyses of single-cell data from macaque PFC have emphasized that neural selectivity is mixed and representations are high dimensional, in seeming contradiction to the findings reported here (Rigotti et al., 2013; Fusi et al., 2016).”
Re-evaluating circuit mechanisms underlying pattern separation (2019, neutral): “Expansion, when combined with nonlinear mixing, increases the dimensionality of the coding subspace (Rigotti et al., 2013; Fusi et al., 2016).”
Comparing continual task learning in minds and machines (2018, background): “…suggested that after extensive (interleaved) training, neurons in prefrontal cortex form high-dimensional representations of task variables that are mixed across the neuronal population (dimension expansion), allowing efficient decoding of relevant actions via appropriate linear readout (45, 46).”
Complementary roles of dimensionality and population structure in neural computations (2020, background): “, 2018], and less prevalent in higher cortical areas where neurons typically multi-task [Raposo et al., 2014; Fusi et al., 2016].”
Towards the neural population doctrine. (2019, background): “As an increasing number of experiments involving multiple categories of stimuli are being performed, neurons in several brain areas have been found to have mixed selectivity: neurons signiﬁcantly modulate their activity in response to more than one stimulus category/type [21,5,19 ].”
Sequential and efficient neural-population coding of complex task information (2019, positive): “Our points above about the theoretical utility of random sequential neural codes extends the literature on how neural circuits can exploit random designs to perform interesting computational functions, such as separable neural representations (Babadi and Sompolinsky, 2014; Fusi et al., 2016; Lindsay”
A designer AAV variant permits efficient retrograde access to projection neurons (2016, background): “378 Neuron 92, 372–382, October 19, 2016 computations are increasingly thought to depend on the dynamics of the entire neuronal population within a particular local circuit module (Fusi et al., 2016).”
The Role of Prefrontal Mixed Selectivity in Cognitive Control (2016, background): “In light of growing evidence for the role of mixed selectivity in an adaptive and dynamic population code (for review, see Fusi et al., 2016), this study opens exciting future avenues.”
Linking Connectivity, Dynamics, and Computations in Low-Rank Recurrent Neural Networks. (2017, neutral): “3, yet it is significantly more complex as it represents a hallmark of cognitive flexibility: a non-linearly separable, XOR-like computation that a single-layer feed-forward network cannot solve (Rigotti et al., 2010; Fusi et al., 2016).”
Continual Learning via Online Leverage Score Sampling (2019, methodology): “Much like our brains, a central strategy employed by our method is to strike a balance between dimensionality reduction of task specific data and dimensionality expansion as processing progresses throughout the hierarchy of the neural network [8].”
If deep learning is the answer, what is the question? (2020, background): “From this perspective, single neurons code for multiple experimental variables and their interactions 38–41 , exhibiting non-linear mixed selectivity.”
The geometry of abstraction in hippocampus and prefrontal cortex (2020, background): “This is because, at the level of individual neurons, the linear component of responses dominates in many situations (Rigotti et al., 2010, 2013; Barak et al., 2013; Lindsay et al., 2017; Fusi et al., 2016), and it is easy to miss the fact that non-linear com-”
Neurodevelopment of the association cortices: Patterns, mechanisms, and implications for psychopathology (2021, positive): “They behave differently (yet reproducibly) depending on the context, thereby enabling adaptable, non-linear, and high-dimensional neural encoding (Fusi et al., 2016; Rigotti et al., 2013).”