0.25
0.5
0.75
1.25
1.5
1.75
2
Correlations and signatures of criticality in neural population models
Published on Mar 07, 20161848 Views
Large-scale recording methods make it possible to measure the statistics of neural population activity, and thereby to gain insights into the principles that govern the collective activity of neural
Related categories
Chapter list
Correlations and signatures of criticality in neural population models00:00
Data00:39
Are neural networks poised at a thermodynamic critical point?01:32
Signatures of criticality in a recording of retinal ganglion cells - 102:08
Signatures of criticality in a recording of retinal ganglion cells - 202:36
Signatures of criticality in a recording of retinal ganglion cells - 302:49
Signatures of criticality in a recording of retinal ganglion cells - 402:58
We consider the distribution P(x) of binary ‘spikewords’ x03:21
‘K-pairwise model’: An Ising model with spike-count constraints04:27
Maximum entropy models can be used to translate concepts from thermodynamics to neural data06:53
Signatures of criticality in a recording of retinal ganglion cells - 509:38
What neural mechanisms can explain this observation?10:25
The simplest possible model of a patch of retina - 112:16
The simplest possible model of a patch of retina - 213:08
We infer model parameters of the K-pairwise model by maximising a penalised log-likelihood13:41
If you use use smart parameter-updates, you do not need to store the entire MCMC sample15:40
Signatures of criticality arise in a simple simulation of the retina, without any tuning or adaptation16:12
Tkacik et al 2015 - our model17:08
Analysis: Subsampling a homogeneous population19:07
In the beta-binomial model, all limits of interest can be calculated in closed form20:37
One source of ‘criticality inducing correlations’23:11
Neural data analysis24:34
Conclusion26:40
Thanks to...27:50