Randomly connected networks: a simple architecture that can drive complex behavior

March 14, 2012

I started thinking about randomly connected networks during a recent visit to the Center for Theoretical Neuroscience where I was a seminar speaker last week. The idea behind these networks, which have been championed recently by Larry Abbott, Stefano Fusi and others, is that neurons in some parts of the brain receive inputs from neurons with a broad array of response properties (see figure, below). This means that, at least from the point of view of an experimenter recording from such neurons, their responses seem at first to be quite confusing. In one paper from Stefano Fusi’s group, they describe prefrontal neurons recorded during a task where monkeys were deciding among a number of different cues, according to a rule that could change over the course of the session. The neurons were recorded in Earl Miller‘s lab at MIT. Each neuron appeared to reflect a complicated combination of the two sensory cues, and the particular context in which they were presented. The paper argues that responses like these are expected in randomly connected networks, and indeed could subserve the very flexibility that is required by tasks that require dynamic combinations on sensory inputs and rules. Might this have implications for multisensory research? We don’t yet know, but like the Miller Lab, we present animals with a complex set of stimuli that vary along several dimensions. A randomly connected network might be the most efficient way for the animals to harvest the relevant information.

From Figure 3 of Rigotti et al

One Response to “Randomly connected networks: a simple architecture that can drive complex behavior”

  1. Daniel said


    I just found your blog and think it’s a great idea for the popularization of research topics (as well as to “randomly connect” with other nodes of our human network 😉 and discussions — nice going. 🙂

    Anyway, i thought this post was very interesting: the idea of using random networks seems quite appropriate, once it can be modeled via Random Matrix Theory and open up a whole box of surprises that way, eg, phase transitions and critical phenomena, and their classification via catastrophe theory. These can possibly lead to a whole set of non-trivial consequences that haven’t quite been fleshed out yet. Very interesting…

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Fairhall lab

Computational neuroscience at the University of Washington

Pillow Lab Blog

Neural Coding and Computation Lab @ Princeton University

Churchland lab

Perceptual decision-making at Cold Spring Harbor

%d bloggers like this: