MENU

Multi-region Neural Network Models of Adaptive and Maladaptive Learning in the Brain

Presented at: Neuroscience 2020
C.E. Credits: P.A.C.E. CE Florida CE
Speaker
  • Assistant Professor, Department of Neuroscience & Friedman Brain Institute, Icahn School of Medicine at Mount Sinai
    Biography
      Dr. Kanaka Rajan is a Computational Neuroscientist and an Assistant Professor of Neuroscience at Mount Sinai. Her research focuses on integrative theories to describe how behavior emerges from the cooperative activity of multi-scale neural processes. These theories use as building blocks, neural network models flexible enough to accommodate various levels of biological detail at the neuronal, synaptic and circuit levels. They are based on high-quality data from neuroscience experiments and on new and existing mathematical and computational frameworks, derived from machine learning and statistical physics, adapted for biological fidelity. She uses a cross-disciplinary approach that provides critical insights into how neural circuits learn and execute functions ranging from working memory to decision making, reasoning and intuition, putting her in a unique position to advance our understanding of how important acts of cognition, such as perception, learning and memory work. For her work in the field of computational neuroscience, she has received a Young Investigator Award from the Brain and Behavior Foundation (formerly, NARSAD), an Understanding Human Cognition Scholar Award from the James S McDonnell Foundation, a Mindlin Foundation Award, two Research Scholars Awards from the Di Sabato Foundation and the Dyal Foundation, and a Sloan Research Fellowship. Her work has been supported by funding from the National Institutes of Health through the BRAIN Initiative and from the National Science Foundation.

    Abstract

    In Andalman et al, 2018, we explored the activity of over 10,000 neurons across more than 15 regions imaged simultaneously in larval zebrafish in a novel behavioral challenge paradigm. Complementing the experimental findings, we developed a computational model of the activity in the habenula and raphe, identifying a significant and specific change in intra-habenular connectivity and in raphe-to-habenula projection strengths as a result of behavioral challenge. These results were consistent with both the key experimental findings of Andalman et al as well as with converging evidence from other studies. Building on this paper, supported by the NIH BRAIN Initiative, for the first time, we consider complex multi-dimensional interactions at the whole brain level, while also relating them causally to observed behavior of the intact organism. We present a novel, more robust, and scalable class of circuit models – multi-region neural network models. These models are constrained at the outset by experimental data at two levels simultaneously: (a) large-scale neural dynamics – cellular resolution activity imaged from all 15 larval zebrafish brain regions – and (b) the behavior of the organism – tail movements tracked realtime in a range of continuous behavioral and experiential states. We develop an innovative method for analyzing and visualizing essential features extracted from our multi-region network models fit directly to these data, and derive from them the measurable signatures of brain-wide fluctuations that are causally predictive of transitions between behavioral states of the fish.

    Our findings demonstrate that theories and mechanistic models constructed in tight conjunction with data, as in this test case, will have far reaching applications for other studies using a range of collection methods and across nervous systems, even those sampled with differing levels of granularity.

    Learning Objectives:

    1. Exemplify the role of principled theory and computational modeling in neuroscience

    2. Present new class of multi-region recurrent network (RNN) models constrained, from the outset, by experimental data

    3. Demonstrate that reverse engineering data-inspired RNN models can uncover key mechanisms of brain-wide inter-area communication that are inaccessible from measurements alone


    Show Resources
    Show Resources