Compare MATLAB and python pRFs

[billbrod]

For some reason, the python and matlab pRFs look like they're different sizes (the images used to create the below plot are shown at the bottom; they're bars sweeping across in the horizontal and vertical directions). Trying to figure out why this is in the visualize_prfs branch on my fork (after I finish with this, I'll open a pull request and then delete my fork, working on branches from this project).

[billbrod]

Alright, so after a bunch of changes, it looks like the only major difference now is that we're truncating the python Gaussian sigmas after a certain number of pRFs, whereas the matlab code doesn't do that.

Another important change: we were casting the pRF_pixel_center as an integer before giving it to matlab (using floor), but that's not necessary; it can handle floats. And using shifted over the pRF slightly. The below all are giving the centers as floats, which you can see cause them to line up much better.

Here's a comparison of predictions with python pRF_blob_stds=2 (blue is python, green is matlab):

and here's with pRF_blob_stds=7:

We can see they're much more similar.

So now the main questions are: what's a reasonable pRF_blob_stds value and can we speed it up? The higher it is, the longer the calculation takes; at 7 it's pretty annoying. I'm having issues running some of my code to completion (it's just taking too long), and I'm not sure why. I'll check it a little more tomorrow, but Noah can you test with the most recent version of the code (on the master trunk) and some different pRF_blob_stds values to see how long it takes? I'm also seeing pRF_blob_stds=2 in my results dictionary even when I've changed it (and, based on the output, it's obviously changed), which is confusing.