Title: Spike Statistics in a HighConductance Cortical Network Model
1Spike Statistics in a High-Conductance Cortical
Network Model
- Joanna Tyrcha
- Stockholm University, Stockholm
-
- in collaboration with John Hertz, Nordita,
Stockholm/Copenhagen
2Outline
Cross-correlations between neurons measured in
network simulations (and experiments)
correlation coefficients 0.17 0.07
3Outline
Cross-correlations between neurons measured in
network simulations (and experiments)
correlation coefficients 0.17 0.07
Finding distribution of spike patterns with the
observed cross-correlations
4Outline
Cross-correlations between neurons measured in
network simulations (and experiments)
correlation coefficients 0.17 0.07
- Finding distribution of spike patterns with the
observed - cross-correlations
- fit with SK spin glass model with ltJijgt gt 0,
-
5Outline
Cross-correlations between neurons measured in
network simulations (and experiments)
correlation coefficients 0.17 0.07
- Finding distribution of spike patterns with the
observed - cross-correlations
- fit with SK spin glass model with ltJijgt gt 0,
- ltJijgt ? 1/N, std(Jij) ?1/N1/6
6Outline
Cross-correlations between neurons measured in
network simulations (and experiments)
correlation coefficients 0.17 0.07
- Finding distribution of spike patterns with the
observed - cross-correlations
- fit with SK spin glass model with ltJijgt gt 0,
- ltJijgt ? 1/N, std(Jij) ?1/N1/6
- in SG phase only for N gt 5000 (?)
7Simulations Model
2 populations in network Excitatory,
Inhibitory
8Simulations Model
2 populations in network Excitatory,
Inhibitory Excitatory external drive
9Simulations Model
2 populations in network Excitatory,
Inhibitory Excitatory external drive HH-like
neurons, conductance-based synapses
10Simulations Model
2 populations in network Excitatory,
Inhibitory Excitatory external drive HH-like
neurons, conductance-based synapses Random
connectivityProbability of connection between
any two neurons is c K/N, where N is the size
of the population and K is the average number of
presynaptic neurons.
11Simulations Model
2 populations in network Excitatory,
Inhibitory Excitatory external drive HH-like
neurons, conductance-based synapses Random
connectivityProbability of connection between
any two neurons is c K/N, where N is the size
of the population and K is the average number of
presynaptic neurons.
Results here for c 0.1, N 1000
12rapidly-varying input
Rext
Stimulus modulation
t (sec)
Flips on and off with exponentially-distributed
on- and off-periods, Mean on/off time 50 ms
13Rasters
inhibitory (100) excitatory (400)
15.5 Hz 8.6 Hz
14Correlation coefficients
Data in 10-ms bins
15Correlation coefficients
Data in 10-ms bins
16Correlation coefficients
Data in 10-ms bins
cc 0.17 0.07
17Modeling the distribution of spike patterns
Have sets of spike patterns Sik Si 1
for spike/no spike (we use 10-ms bins) (temporal
order irrelevant)
18Modeling the distribution of spike patterns
Have sets of spike patterns Sik Si 1
for spike/no spike (we use 10-ms bins) (temporal
order irrelevant) Construct a distribution PS
that generates the observed patterns (i.e., has
the same correlations)
19Modeling the distribution of spike patterns
Have sets of spike patterns Sik Si 1
for spike/no spike (we use 10-ms bins) (temporal
order irrelevant) Construct a distribution PS
that generates the observed patterns (i.e., has
the same correlations) Simplest nontrivial
model (Schneidman et al, Nature 440 1007 (2006),
Tkacik et al, arXivq-bio.NC/0611072)
20Modeling the distribution of spike patterns
Have sets of spike patterns Sik Si 1
for spike/no spike (we use 10-ms bins) (temporal
order irrelevant) Construct a distribution PS
that generates the observed patterns (i.e., has
the same correlations) Simplest nontrivial
model (Schneidman et al, Nature 440 1007 (2006),
Tkacik et al, arXivq-bio.NC/0611072)
parametrized by Jij, hi
21Fitting parameters of the distribution (I)
Boltzmann learning
Iterative adjustment of Jij, hi
22Fitting parameters of the distribution (I)
Boltzmann learning
Iterative adjustment of Jij, hi
23Fitting parameters of the distribution (I)
Boltzmann learning
Iterative adjustment of Jij, hi
requires knowing only first- and second-order
spike statistics
24Fitting parameters of the distribution (I)
Boltzmann learning
Iterative adjustment of Jij, hi
requires knowing only first- and second-order
spike statistics A model like would need
3rd-order statistics to find
25Fitting parameters of the distribution (II)
inversion of Thouless-Anderson-Palmer equations
(Tanaka, PRE 58 2302 (1998))
26Fitting parameters of the distribution (II)
inversion of Thouless-Anderson-Palmer equations
(Tanaka, PRE 58 2302 (1998)) TAP free
energy
27Fitting parameters of the distribution (II)
inversion of Thouless-Anderson-Palmer equations
(Tanaka, PRE 58 2302 (1998)) TAP free
energy ?F/?mi 0 gt TAP equations
28Fitting parameters of the distribution (II)
inversion of Thouless-Anderson-Palmer equations
(Tanaka, PRE 58 2302 (1998)) TAP free
energy ?F/?mi 0 gt TAP equations
29TAP inversion (continued) Tanaka procedure
- Measure mi ltSigt and Cij ltSiSjgt - mimj from
data
30TAP inversion (continued) Tanaka procedure
- Measure mi ltSigt and Cij ltSiSjgt - mimj from
data - Invert C and solve for Jij in
31TAP inversion (continued) Tanaka procedure
- Measure mi ltSigt and Cij ltSiSjgt - mimj from
data - Invert C and solve for Jij in
- Solve TAP equations for hi
32J distributions
33J distributions
cf Tkacik et al
34Mean J ?1/N
35Standard deviationsfall off slower than 1/N1/2
36Sherrington-Kirkpatrick spin glass model
with
37Sherrington-Kirkpatrick spin glass model
with
Infinite-range (mean field) model
38Sherrington-Kirkpatrick spin glass model
with
Infinite-range (mean field) model
Here,
39J of equivalent SK model increases with N
40Heading for a spin glass state?
Almeida and Thouless, J Phys A 1978
41Heading for a spin glass state?
here at N80
Almeida and Thouless, J Phys A 1978
42Heading for a spin glass state?
here at N80
here at N160
Almeida and Thouless, J Phys A 1978
43Heading for a spin glass state?
here at N80
here at N160
reach SG at N5000
Almeida and Thouless, J Phys A 1978
44J distributions are not normal
45Perspective
- Fitting Ising models for PS results look like
SK models -
-
-
-
-
46Perspective
- Fitting Ising models for PS results look like
SK models - -- 2 methods Boltzmann learning and inversion
of TAP equations -
-
-
-
47Perspective
- Fitting Ising models for PS results look like
SK models - -- 2 methods Boltzmann learning and inversion
of TAP equations - -- J0 independent of N, J ? N1/3
-
-
-
48Perspective
- Fitting Ising models for PS results look like
SK models - -- 2 methods Boltzmann learning and inversion
of TAP equations - -- J0 independent of N, J ? N1/3
- -- higher-order couplings not necessary
-
-
49Perspective
- Fitting Ising models for PS results look like
SK models - -- 2 methods Boltzmann learning and inversion
of TAP equations - -- J0 independent of N, J ? N1/3
- -- higher-order couplings not necessary
-
50Perspective
- Fitting Ising models for PS results look like
SK models - -- 2 methods Boltzmann learning and inversion
of TAP equations - -- J0 independent of N, J ? N1/3
- -- higher-order couplings not necessary
- -- no spin glass phase for Nlt5000.