BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Computational Neuroscience Journal Club - Will Greedy\, University
  of Bristol
DTSTART:20230530T130000Z
DTEND:20230530T150000Z
UID:TALK200770@talks.cam.ac.uk
CONTACT:Jake Stroud
DESCRIPTION:Please join us for our fortnightly Computational Neuroscience 
 journal club on Tuesday 30th May at 2pm UK time in the CBL seminar room\, 
 or online on zoom. The title is ‘Dendritic computations and backpropagat
 ion’\, presented by Will Greedy from the University of Bristol.\n\nhttps
 ://eng-cam.zoom.us/j/84204498431?pwd=Um1oU284b1YxWThObGw4ZU9XZitWdz09\nMee
 ting ID: 842 0449 8431 Passcode: 684140\n\nSummary:\nThe error-backpropaga
 tion (backprop) algorithm remains the most common solution to the credit a
 ssignment problem in artificial neural networks. In neuroscience\, it is u
 nclear whether the brain could adopt a similar strategy to correctly modif
 y its synapses. Recent models have attempted to bridge this gap while bein
 g consistent with a range of experimental observations.\n \nIn this journa
 l club\, I will:\n \n1. Introduce the general field of backprop in dendrit
 ic networks.\n\n2. Go over Error-encoding Dendritic Networks (EDNs\; Sacra
 mento et al. 2018 NeurIPS). In this paper\, it was first introduced the id
 ea of using cell-types and distal dendrites to jointly encode error signal
 s.\n\n3. Next\, I will introduce Burstprop (Payeur et al. Nature Neuro 202
 1)\, which proposes that there are two types of signals in the brain. Sing
 le-spike events for inference and bursts for learning. This then suggests 
 the need for specialised short-term synaptic plasticity with which to deco
 de these signals.\n\n4. Both EDNs and burstprop are either unable to effec
 tively backpropagate error signals across multiple layers or require a mul
 ti-phase learning process\, neither of which are reminiscent of learning i
 n the brain. Next\, I will introduce our recent model\, Bursting Cortico-C
 ortical Networks (BurstCCN\; Greedy et al. Neurips 2022)\, which solves th
 ese issues by integrating known properties of cortical networks namely bur
 sting activity\, short-term plasticity (STP) and dendrite-targeting intern
 eurons.\n \nOverall\, these results suggest that cortical features across 
 sub-cellular\, cellular\, microcircuit\, and systems levels jointly underl
 ie single-phase efficient deep learning in the brain.
LOCATION:In Person (CBL Seminar Room) and Online on Zoom
END:VEVENT
END:VCALENDAR