BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:LCN Seminar: Guillaume Bellec
DTSTART:20190618T140000
DTEND:20190618T150000
DTSTAMP:20260929T050730Z
UID:323ecfc2142f8d397ecf1db603497be197f7233eb15771eb4f8679a3
CATEGORIES:Conferences - Seminars
DESCRIPTION:Biologically inspired alternatives to backpropagation through 
 time for learning in recurrent neural networks\n\nThe way how recurrently 
 connected networks of spiking neurons in the brain acquire powerful inform
 ation processing capabilities through learning has remained a mystery. Thi
 s lack of understanding is linked to a lack of learning algorithms for rec
 urrent networks of spiking neurons (RSNNs) that are both functionally powe
 rful and can be implemented by known biological mechanisms. Since RSNNs ar
 e simultaneously a primary target for implementations of brain-inspired ci
 rcuits in neuromorphic hardware\, this lack of algorithmic insight also hi
 nders technological progress in that area. The gold standard for learning 
 in recurrent neural networks in machine learning is back-propagation throu
 gh time (BPTT)\, which implements stochastic gradient descent with regard 
 to a given loss function. But BPTT is unrealistic from a biological perspe
 ctive\, since it requires a transmission of error signals backwards in tim
 e and in space\, i.e.\, from post- to presynaptic neurons. We show that an
  online merging of locally available information during a computation with
  suitable top-down learning signals in real-time provides highly capable a
 pproximations to BPTT. For tasks where information on errors arises only l
 ate during a network computation\, we enrich locally available information
  through feedforward eligibility traces of synapses that can easily be com
 puted in an online manner. The resulting new generation of learning algori
 thms for recurrent neural networks provides a new understanding of network
  learning in the brain that can be tested experimentally. In addition\, th
 ese algorithms provide efficient methods for on-chip training of RSNNs in 
 neuromorphic hardware.
LOCATION:AAC 0 06 https://plan.epfl.ch/?room=AAC006
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR