BioSpaun: A large-scale behaving brain model with complex neurons


#1

What makes Spaun unique among large-scale brain models is its functional
abilities. Spaun receives input from the environment through its single
eye, which is shown images of handwritten or typed digits and letters,
and it manipulates the environment by moving a physically modeled arm,
which has mass, length, inertia, and so on. Spaun uses these natural interfaces,
in combination with internal cognitive processes, to perceive visual
input, remember information, reason using that information, and generate
motor output (writing out numbers or letters). It uses these abilities to
perform eight different tasks, ranging from perceptual-motor tasks (recreating
the appearance of a perceived digit) to reinforcement learning (in a
gambling task) to language-like inductive reasoning (completing abstract
patterns in observed sequences of digits). These tasks can be performed
in any order, they are all executed by the same model, and there are no
changes to the model between tasks. To see the model perform the tasks,
see http://nengo.ca/build-a-brain/spaunvideos.

In this paper we incorporate detailed compartmental models of the type
used in the recent HBP model into different cortical areas of our large-scale,
behaving brain model. We refer to this augmented model as “BioSpaun”. We
show that the behavior of the original Spaun model is not adversely affected
by changing the neuron model. We further show that the additional complexity
can be used to test hypotheses not possible with the original model.
Specifically, we demonstrate that BioSpaun can be used to simulate the effects
of adding the drug tetrodotoxin (TTX) to these areas of cortex. We
perform this manipulation to both visual cortex and frontal cortex, demonstrating
performance declines related to the dosage of drug applied, both
within and across different tasks. While much remains to be done to verify
the accuracy of these simulations in vivo, we believe this is the first demonstration
of a large-scale behaving neural model that includes a high degree
of biophysical detail. Integrating these two aspects of brain modeling provides
a new method for testing low-level molecular and other physiological
interventions on high-level behavior.

https://arxiv.org/abs/1602.05220