Biology subjects

Xiao-Jing Wang

Publications and source records attributed to Xiao-Jing Wang.

5 recordsLinked to original sources

bioRxiv2016

Reward-based training of recurrent neural networks for cognitive and value-based tasks

Trained neural network models, which exhibit many features observed in neural recordings from behaving animals and whose activity and connectivity can be fully analyzed, may provide insights into neural mechanisms. In contrast to commonly used methods for supervised learning from graded error signals, however, animals learn from reward feedback on definite actions through reinforcement learning. Reward maximization is particularly relevant when the optimal behavior depends on an animals internal judgment of confidence or subjective preferences. Here, we describe reward-based training of recurrent neural networks in which a value network guides learning by using the selected actions and activity of the policy network to predict future reward. We show that such models capture both behavioral and electrophysiological findings from well-known experimental paradigms. Our results provide a unified framework for investigating diverse cognitive and value-based computations, including a role for value representation that is essential for learning, but not executing, a task.

Neuroscience↗

bioRxiv2016

Feedforward and feedback frequency-dependent interactions in a large-scale laminar network of the primate cortex

Interactions between top-down and bottom-up processes in the cerebral cortex hold the key to understanding predictive coding, executive control and a gamut of other brain functions. The underlying circuit mechanism, however, remains poorly understood and represents a major challenge in neuroscience. In the present work we tackled this problem using a large-scale computational model of the primate cortex constrained by new directed and weighted connectivity data. In our model, the interplay between feedforward and feedback signaling depends on the cortical laminar structure and involves complex dynamics across multiple (intra-laminar, inter-laminar, inter-areal and whole cortex) scales. The model was tested by reproducing, and shedding insights into, a wide range of neurophysiological findings about frequency-dependent interactions between visual cortical areas: feedforward pathways are associated with enhanced gamma (30-70 Hz) oscillations, whereas feedback projections selectively modulate alpha/low beta (8-15 Hz) oscillations. We found that in order for the model to account for the experimental observations, the feedback projection needs to predominantly target infragranular layers in a target area, which leads to a proposed circuit substrate for predictive coding. The model reproduces a functional hierarchy based on frequency-dependent Granger causality analysis of inter-areal signaling, as reported in recent monkey and human experiments. Taken together, this work highlights the importance of multi-scale approaches and provides a modeling platform for studies of large-scale brain circuit dynamics and functions.

Neuroscience↗

bioRxiv2016

A Dendritic Disinhibitory Circuit Mechanism for Pathway-Specific Gating

In this work we propose that a disinhibitory circuit motif, which recently gained experimental support, can instantiate flexible routing of information flow along selective pathways in a complex system of cortical areas according to behavioral demands (pathway-specific gating). We developed a network model of pyramidal neurons and three classes of interneurons, with connection probabilities constrained by data. If distinct input pathways cluster on separate dendritic branches of pyramidal neurons, then a pathway can be gated-on by disinhibiting targeted dendrites. We show that this branch-specific disinhibition can be achieved despite dense in-terneuronal connectivity, even under the assumption of random connections. We found clustering of input pathways on dendrites can emerge through synaptic plasticity regulated by disinhibition. This gating mechanism in a neural circuit is further demonstrated by performing a context-dependent decision-making task. Our findings suggest a microcircuit architecture that harnesses dendritic computation and diverse inhibitory neuron types to subserve cognitive flexibility.

Neuroscience↗

bioRxiv2016

Random recurrent networks near criticality capture the broadband power distribution of human ECoG dynamics

The power spectrum of brain electric field potential recordings is dominated by an arrhythmic broadband signal but a mechanistic account of its underlying neural network dynamics is lacking. Here we show how the broadband power spectrum of field potential recordings can be explained by a simple random network of nodes near criticality. Such a recurrent network produces activity with a combination of a fast and a slow autocorrelation time constant, with the fast mode corresponding to local dynamics and the slow mode resulting from recurrent excitatory connections across the network. These modes are combined to produce a power spectrum similar to that observed in human intracranial EEG (i.e., electrocorticography, ECoG) recordings. Moreover, such a network naturally converts input correlations across nodes into temporal autocorrelation of the network activity. Consequently, increased independence between nodes results in a reduction in low-frequency power, which offers a possible explanation for observed changes in ECoG power spectra during task performance. Lastly, changes in network coupling produce changes in network activity power spectra reminiscent of those seen in human ECoG recordings across different arousal states. This model thus links macroscopic features of the empirical ECoG power spectrum to a parsimonious underlying network structure and proposes potential mechanisms for changes in ECoG power spectra observed across behavioral and arousal states. This provides a computational framework within which to generate and test hypotheses about the cellular and network mechanisms underlying whole brain electrical dynamics, their variations across behavioral states as well as abnormalities associated with brain diseases.

Neuroscience↗

bioRxiv2015

A large-scale circuit mechanism for hierarchical dynamical processing in the primate cortex

We developed a large-scale dynamical model of the macaque neocortex based on recent quantitative connectivity data. A hierarchy of timescales naturally emerges from this system: sensory areas show brief, transient responses to input (appropriate for sensory processing), whereas association areas integrate inputs over time and exhibit persistent activity (suitable for decision-making and working memory). The model displays multiple temporal hierarchies, as evidenced by contrasting responses to visual and somatosensory stimulation. Moreover, slower prefrontal and temporal areas have a disproportionate impact on global brain dynamics. These findings establish for the first time a circuit mechanism for \"temporal receptive windows\" that are progressively enlarged along the cortical hierarchy, extend the concept of working memory from local to large circuits, and suggest a re-interpretation of functional connectivity measures.\n\nConflict of InterestThe authors declare no competing financial interests.\n\nAcknowledgementsWe thank Nikola Markov and John Murray for discussions. This work was supported by ONR grant N00014-13-1-0297 and NIH grant R01MH062349 (X.-J.W.) and by ANR-11-BSV4-501 and LABEX CORTEX (ANR-11-LABX-0042) of Universite de Lyon, program \"Investissements dAvenir\" (ANR-11-IDEX-0007) operated by the French National Research Agency (H.K.). The connectivity data is available at www.core-nets.org.

Neuroscience↗