Santa Fe Institute Collaboration Platform

COMPLEX TIME: Adaptation, Aging, & Arrow of Time

Get Involved!
Contact: Caitlin Lorraine McShea, Program Manager, cmcshea@santafe.edu

Cognitive Regime Shift II - When/why/how the Brain Breaks/Task-performing neural network models enable us to test theories of brain computation with brain and behavioral data

From Complex Time

November 12, 2019
11:30 am - 12:30 pm

Presenter

Nikolaus Kriegeskorte (Columbia Univ.)

Abstract

The goal of computational neuroscience is to find mechanistic explanations of how the nervous system processes information to support cognitive function and behavior. Deep neural networks (DNNs), using feedforward or recurrent architectures, have come to dominate several domains of artificial intelligence (AI). As the term “neural network” suggests, these models are inspired by biological brains. However, their units are rate-coded linear-nonlinear elements, abstracting from the intricacies of biological neurons, including their spatial structure, ion channels, and complex dentritic and axonal signalling dynamics. The abstractions enable DNNs to be efficiently implemented in computers, so as to perform complex feats of intelligence, ranging from perceptual tasks (e.g. visual object and auditory speech recognition) to cognitive tasks (e.g. language translation), and on to motor control tasks (e.g. playing computer games or controlling a robot arm). In addition to their ability to model complex intelligent behaviors, DNNs have been shown to predict neural responses to novel sensory stimuli that cannot be predicted with any other currently available type of model. DNNs can have millions of parameters (connection strengths), which are required to capture the domain knowledge needed for task performance. These parameters are often set by task training using stochastic gradient descent. The computational properties of the units are the result of four directly manipulable elements: (1) functional objective, (2) network architecture, (3) learning algorithm, and (4) input statistics. The advances with neural nets in engineering provide the technological basis for building task-performing models of varying degrees of biological realism that promise substantial insights for computational neuroscience.  

Presentation file(s)
Related files