Lecture, 21.01.19 biophysics
  • HH nullclines, linear scale
  • HH nullclines, log scale
    • Lecture, 21.01.22 biophysics; a writeup covering part of the first lecture is here.
    Lecture, 21.01.26 biophysics
    Lecture, 21.01.29 biophysics
    Lecture, 21.02.02 random networks
    Lecture, 21.02.05 random networks
    Lecture, 21.02.09 random and structured networks
    Lecture, 21.02.12 structure: Hopfield and line attractor networks
    Lecture, 21.02.16 line attractor networks and general recurrent neural networks
    Lecture, 21.02.19 mainly backprop through time
    Lecture, 21.02.23 mainly backprop, and grand summary

    Lecture notes:
  • You can find previous tests and lecture notes here.
    • More specifically,
  • biophysics.
  • linear analysis and linear algebra.
  • randomly connected networks.
  • Hopfield networks.
  • line attractor networks.
  • networks with tine-varying dynamics, including learning.
  • dynamic mean field analysis, based on this paper.


    • Relevant papers
    single neuron dynamics:
  • Demonstration that all type I neurons reduce to the same set of equations at low firing rate.
    • synaptic plasiticity:
  • Graupner and Brunel's model of synaptic plasticity.
  • Mark van Rossum's model for stable STDP.
    • cascade models of learning:
  • Fusi, Drew and Abbott 2005.
  • Benna and Fusi 2016.
    • randomly connected networks
  • van Vreeswijk and Sompolinsky's classic paper. hard, but thorough.
  • firing rate dynamics -- possibly most useful for the construction of nullclines.
    • Hopfield networks
  • Hopfield 1982
  • Hopfield 1984
    • More realistic Hopfield networks
  • Latham and Nirenberg 2004.
  • Roudi and Latham 2007
    • Derivation of backprop through time (in the discrete time setting)
  • James Murray, Local online learning in recurrent networks with random feedback (2019)
    • Biologically plausible (ish) learning rules
  • Feedback alignment (2016)
  • Direct feedback alignment (2016)
  • Feedback alignment with learning (2020)
  • Bottleneck method, but not exactly biologicallly plausible (2019)
  • Bottleneck method, but more biologicallly plausible (2020)
  • Gated linear networks (2017). There are now seveal papers on these networks; this was the original.