Cookies
We use essential cookies to run the site. Analytics cookies are optional and help us improve World Wide. Learn more.
Institute of Neuroinformatics, University of Zürich and ETH Zürich
Showing your local timezone
Schedule
Thursday, September 30, 2021
2:00 PM Europe/London
Recording provided by the organiser.
Domain
NeuroscienceHost
Sheffield ML
Duration
70 minutes
The success of deep learning sparked interest in whether the brain learns by using similar techniques for assigning credit to each synaptic weight for its contribution to the network output. However, the majority of current attempts at biologically-plausible learning methods are either non-local in time, require highly specific connectivity motives, or have no clear link to any known mathematical optimization method. Here, we introduce Deep Feedback Control (DFC), a new learning method that uses a feedback controller to drive a deep neural network to match a desired output target and whose control signal can be used for credit assignment. The resulting learning rule is fully local in space and time and approximates Gauss-Newton optimization for a wide range of feedback connectivity patterns. To further underline its biological plausibility, we relate DFC to a multi-compartment model of cortical pyramidal neurons with a local voltage-dependent synaptic plasticity rule, consistent with recent theories of dendritic processing. By combining dynamical system theory with mathematical optimization theory, we provide a strong theoretical foundation for DFC that we corroborate with detailed results on toy experiments and standard computer-vision benchmarks.
Alexander Meulemans
Institute of Neuroinformatics, University of Zürich and ETH Zürich
Contact & Resources
neuro
Nervous system function relies on the polarized architecture of neurons, established by directional transport of pre- and postsynaptic cargoes. While delivery of postsynaptic components depends on the
neuro
A combinatorial neural code for long-term motor memory
neuro
The postdoctoral researcher will work on an interdisciplinary project at the intersection of deep learning and comparative politics. The candidate will work in the Human-Centered Machine Learning (HuM
We are looking for highly motivated researchers to join our group in interdisciplinary projects that focus on the development of computational models to understand how linguistic information is repres
The CSNG Lab at the Faculty of Mathematics and Physics at the Charles University is seeking a highly motivated Postdoctoral Researcher to join our team to work on a digital twin model of the visual sy
PROJECT SUMMARY / ABSTRACT Cluster headache (CH), which affects about 1 in 1,000 people, is a severe and debilitating primary headache disorder characterized by repeated attacks occurring in clusters