tag:blogger.com,1999:blog-2878020709809350448.post1325089837839425720..comments2017-11-30T09:35:30.794-05:00Comments on ...And You Will Know Me By The Trail of Papers: Compendium of Analyses, Part II, EnsemblesMikehttp://www.blogger.com/profile/13460212509225238067noreply@blogger.comBlogger3125tag:blogger.com,1999:blog-2878020709809350448.post-16219893640329529222011-10-15T15:47:22.088-04:002011-10-15T15:47:22.088-04:00Wow, comments!
Re: cross-correlations, I posted a...Wow, comments!<br /><br />Re: cross-correlations, I posted about this on Google+, and Tim Hanson suggested the same histogram trick you did, and Stephen Shepherd recommended Chronux. In the end I found this script which does it:<br />http://www.weizmann.ac.il/home/bndmitri/MatlabTutorialOnline/Matlab_tut_v4.htm. The run time is obviously worse for large data sets, but it's quick with what I&#Mikehttps://www.blogger.com/profile/13460212509225238067noreply@blogger.comtag:blogger.com,1999:blog-2878020709809350448.post-24028426444368678692011-10-03T13:35:35.314-04:002011-10-03T13:35:35.314-04:00Concerning the HMM, if you're too lazy to impl...Concerning the HMM, if you're too lazy to implement vector output, you can use a standard HMM with scalar output with N = number of neurons + 1 symbols. Use small bins (say, .5ms) and encode the vector output at time t with scalar symbols as follows:0 -> no spike, 1 -> spike from neuron 1, 2-> spike from neuron 2 etc. In the off chance there's two spikes in a single time step Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-2878020709809350448.post-64368763773976610052011-10-03T10:23:57.628-04:002011-10-03T10:23:57.628-04:00Xcorr should work fine for 'digital' signa...Xcorr should work fine for 'digital' signals. After all, if you are working in Matlab, you are never working with a truly analog representation of the data (unless you are doing your analysis on symbolic representations, which would be interesting but unusual). We give Matlab our sampled data, which is digital. <br /><br />So I think you mean it won't apply for data represented as a Eric Thomsonhttps://www.blogger.com/profile/06847717704454032165noreply@blogger.com