A canonical brain computation: from mechanism to purpose

Add to your list(s) Download to your calendar using vCal

• Dr. Yashar Ahmadian, University of Cambridge
• Thursday 26 November 2020, 14:00-15:00
• Online (Zoom).

If you have a question about this talk, please contact Thiago Burghi.

Zoom meeting link: https://zoom.us/j/99373703786

To support global perception, neural networks in the cerebral cortex have to integrate information across local stimulus features. Such integration manifests, for example, in “normalization”: the sublinear summation of responses to combinations of local features. Normalization is performed across cortical areas and is considered a canonical brain computation. While normalization is typically sub-additive and suppressive, its sublinearity weakens with diminishing stimulus strength. I will start by reviewing a parsimonious model of cortical circuitry which mechanistically explains this weakening, and moreover predicts a transition to facilitative and super-additive multi-feature integration for weak stimuli. I will then present a normative theory of this transition in the case of the primary visual cortex. I will show how the notion of optimal coding of natural scenes, in the presence of biological noise, robustly predicts the same transition from sub- to super-additive multi-input integration.

Zoom link: https://zoom.us/j/99373703786

This talk is part of the CUED Control Group Seminars series.

This talk is included in these lists:

• All Talks (aka the CURE list)
• CUED Control Group Seminars
• Cambridge Centre for Data-Driven Discovery (C2D3)
• Cambridge University Engineering Department Talks
• Cambridge talks
• Centre for Smart Infrastructure & Construction
• Chris Davis' list
• Computational Continuum Mechanics Group Seminars
• Featured lists
• Information Engineering Division seminar list
• Interested Talks
• Online (Zoom)
• School of Technology
• Signal Processing and Communications Lab Seminars
• Trust & Technology Initiative - interesting events
• bld31
• ndk22's list
• ob366-ai4er
• rp587

Note that ex-directory lists are not shown.