A geometric framework for understanding dynamic information integration in context-dependent computation

Abstract

Summary The prefrontal cortex (PFC) plays a prominent role in performing flexible cognitive functions and working memory, yet the underlying computational principle remains poorly understood. Here, we trained a rate-based recurrent neural network (RNN) to explore how the context rules are encoded, maintained across seconds-long mnemonic delay, and subsequently used in a context-dependent decision-making task. The trained networks replicated key experimentally observed features in the PFC of rodent and monkey experiments, such as mixed selectivity, neuronal sequential activity, and rotation dynamics. To uncover the high-dimensional neural dynamical system, we further proposed a geometric framework to quantify and visualize population coding and sensory integration in a temporally defined manner. We employed dynamic epoch-wise principal component analysis (PCA) to define multiple task-specific subspaces and task-related axes, and computed the angles between task-related axes and these subspaces. In low-dimensional neural representations, the trained RNN first encoded the context cues in a cue-specific subspace, and then maintained the cue information with a stable low-activity state persisting during the delay epoch, and further formed line attractors for sensor integration through low-dimensional neural trajectories to guide decision-making. We demonstrated via intensive computer simulations that the geometric manifolds encoding the context information were robust to varying degrees of weight perturbation in both space and time. Overall, our analysis framework provides clear geometric interpretations and quantification of information coding, maintenance, and integration, yielding new insight into the computational mechanisms of context-dependent computation.

Mediodorsal thalamus regulates sensory and mapping uncertainties in flexible decision making

“…Fixed-point analysis. Similar to our previous analysis 21,22 , we identified fixed-points or slow points of the performance-optimized PFC-MD model by numerically solving the optimization problem (https://github.com/mattgolub/fixed-point-finder) min 𝐱 j 𝑞(𝐱 j), where 𝑞(𝐱 j) = l−𝐱 j + 𝐖 ,== 𝜙(𝐱 j) + [𝐖 /0 , 𝐕 /0 ] • 𝐮l p (which was not certified by peer review) is the author/funder. All rights reserved.…”

“…4a-f ). From a neural trajectory p ( t ), we further defined the time-varying neural velocity by computing the change rate in neural trajectory 21 : v ( t ) = ║ṗ( t )║ ( Fig. 4g ).…”

“…In addition, how the newly discovered cellular diversity in MD thalamus contributes to such computations is unexplored. Biologically-inspired computational modeling provides a powerful approach to probe important questions that yield mechanistic insight into neural circuits [18][19][20][21][22][23][24][25] . Development of computational circuit models that incorporate neuronal subtype and circuit pathway knowledge is vital in this direction [26][27][28] .Motivated by accumulating data in rodent and human experiments 4,13,14,16,17 , here we develop biological constrained models for the PFC-MD network with cortical interneuron subtype and thalamocortical pathway specificity.…”

Excitatory-Inhibitory Recurrent Dynamics Produce Robust Visual Grids and Stable Attractors

“…Our demonstration of visual grid patterns produced from a CNN-RNN model suggests that the computational principles of grid computation is beyond the velocity-driven path integration task. The RNN has provided a dynamical systems viewpoint for motor movement (Sussilo et al 2015; Michaels et al 2016), path integration (Cueva and Wei 2018; Sorscher et al 2020), information integration (Zhang et al 2021), and predictive representations (Recanatesi et al 2021). We envision that the recurrent dynamics of RNNs can be biologically implemented by a neural substrate in a wide range of cortical networks outside the traditional hippocampus-entorhinal system.…”

Spiking Recurrent Neural Networks Represent Task-Relevant Neural Sequences in Rule-Dependent Computation

“…A biological brain is large-scale neuronal network with recurrent connections that performs computations in complex task behaviors. In recent years, recurrent neural networks (RNNs) have been widely used for modeling a wide range of neural circuits, such as the prefrontal cortex (PFC), parietal cortex, and primary motor cortex (M1), in various cognitive and motor tasks [1][2][3][4][5][6][7][8][9]. Varying types of assumptions have been made in different models: leaky integrate-and-fire (LIF) vs. conductance-based compartment model, rate vs. spiking-based model.…”