Computing with spikes Romain Brette Ecole Normale Supérieure.

download Computing with spikes Romain Brette Ecole Normale Supérieure.

If you can't read please download the document

Transcript of Computing with spikes Romain Brette Ecole Normale Supérieure.

  • Slide 1
  • Computing with spikes Romain Brette Ecole Normale Suprieure
  • Slide 2
  • Spiking neuron models
  • Slide 3
  • Input = N spike trains Output = 1 spike train What transformation ?
  • Slide 4
  • Synaptic integration spike threshold action potential or spike postsynaptic potential (PSP) temporal integration Integrate and fire model: spikes are produced when the membrane potential exceeds a threshold
  • Slide 5
  • The membrane Lipid bilayer (= capacitance) with pores (channels = proteins) outside Na + Cl - inside K + 2 nm specific capacitance 1 F/cm total specific capacitance = specific capacitance * area
  • Slide 6
  • The resting potential At rest, the neuron is polarized: V m -70 mV The membrane is semi-permeable Mostly permeable to K+ A potential difference appears and opposes diffusion V Membrane potential V m =V in -V out K+ diffusion
  • Slide 7
  • Electrodiffusion Membrane permeable to K+ only Diffusion creates an electrical field Electrical field opposes diffusion Equilibrium potential or Nernst potential : or reversal potential F = 96 000 C.mol -1 (Faraday constant) R = 8.314 J.K -1.mol -1 (gas constant) z = charge of ion extra intra
  • Slide 8
  • The equivalent electrical circuit I Linear approximation of leak current: I = g L (V m -E L ) leak or resting potential leak conductance = 1/Rmembrane resistance = capacitance E L -70 mV : the membrane is polarized (V in < V out )
  • Slide 9
  • The membrane equation I inj outside inside =1/R VmVm membrane time constant (typically 3-100 ms)
  • Slide 10
  • Synaptic currents synaptic current postsynaptic neuron synapse I s (t)
  • Slide 11
  • Postsynaptic potentials (PSPs) Presynaptic neuron (extracellular electrode) Postsynaptic neuron (intracellular electrode) (cat motoneuron)
  • Slide 12
  • Idealized synapse Total charge Opens for a short duration I s (t)=Q (t) Dirac function ELEL Spike-based notation: at t=0 w=RQ/ is the synaptic weight
  • Slide 13
  • Temporal and spatial integration Response to a set of spikes {t i j } ? Linearity: i = synapse j = spike number Superposition principle
  • Slide 14
  • Synaptic integration and spike threshold spike threshold action potential PSP Integrate-and-fire : If V = V t (threshold) then: neuron spikes and V V r (reset)
  • Slide 15
  • The integrate-and-fire model Differential formulation spike at synapse i Integral formulation If V = V t (threshold) then: neuron spikes and V V r (reset)
  • Slide 16
  • Is this a sensible description of neurons?
  • Slide 17
  • A phenomenological approach Injected current Recording Model (IF with adaptive threshold fitted with Brian + GPU) Fitting spiking models to electrophysiological recordings Rossant et al. (Front. Neuroinform. 2010)
  • Slide 18
  • Results: regular spiking cell Winner of INCF competition: 76% (variation of adaptive threshold IF) Rossant et al. (2010). Automatic fitting of spiking neuron models to electrophysiological recordings (Frontiers in Neuroinformatics)
  • Slide 19
  • Good news Adaptive integrate-and-fire models are excellent phenomenological models! (response to somatic injection)
  • Slide 20
  • Advanced concepts Synaptic channels are also described by electrodiffusion Neurons are not isopotential ionic channel conductance synaptic reversal potential g s (t) presynaptic spike open closed The cable equation
  • Slide 21
  • Simulating spiking models with PiPe Ce Ci P from brian import * eqs=''' dv/dt = (ge+gi-(v+49*mV))/(20*ms) : volt dge/dt = -ge/(5*ms) : volt dgi/dt = -gi/(10*ms) : volt '' P=NeuronGroup(4000,model=eqs, threshold=v>-50*mV,reset=v=-60*mV) P.v=-60*mV+10*mV*rand(len(P)) Pe=P.subgroup(3200) Pi=P.subgroup(800) Ce=Connection(Pe,P,'ge',weight=1.62*mV,sparseness=0.02) Ci=Connection(Pi,P,'gi',weight=-9*mV,sparseness=0.02) M=SpikeMonitor(P) run(1*second) raster_plot(M) show() briansimulator.org Goodman, D. and R. Brette (2009). The Brian simulator. Front Neurosci doi:10.3389/neuro.01.026.2009.
  • Slide 22
  • Computing with spikes I: rate-based theories
  • Slide 23
  • Statistics of spike trains Spike train: A sequence of spike times (t k ) A signal Rate: Number of spikes / time (= lim k/t k ) Average of S: t1t1 t2t2 t3t3 t n+1 t n = interspike interval (ISI) (Up to a few hundred Hz)
  • Slide 24
  • Rate-based descriptions I s (t) I s (t) = total synaptic current F F1F1 F2F2 F3F3 F4F4 Rate F Can we express F as a function of F 1, F 2,..., F n ? Inputs with rates F 1, F 2,..., F n
  • Slide 25
  • Approach #1: the perfect integrator Neglect the leak current: More precise: replace V m by
  • Slide 26
  • The perfect integrator Normalization V t =1, V r =0 et si V=V t : V V r = change of variable for V: V* = (V-V r )/(V t -V r ) Same for I
  • Slide 27
  • The perfect integrator Integration: Firing rate: if otherwise Hz Brette, R. (2004). Dynamics of one-dimensional spiking neuron models. J Math Biol
  • Slide 28
  • The perfect integrator with synaptic inputs J k = postsynaptic current (superposition principle) timing of spike at synapse k (= 0 for t