What is what in PVC-2?

Hi community

My name is Ilya, I am a PhD student at the Computational Neuroscience lab in University of Tartu and I have a few questions about the PVC-2 dataset.

I am interested in the part of the dataset related to the natural images (<pre>crcns-pvc2/2D_noise_natural</pre> and I have a trouble understanding the structure of the data​:

1) In the articles it is mentioned that "three distinct natural ensembles were used in this study", but there are 8 <pre>Equalpower*</pre> files inside the <pre>Stimulus_Files</pre>. Why is that? What letters A,B,C and numbers 1,2,3 mean? Are these different sequences of images or same images but in different order?

2) Inside the <pre>Spike_and_Log_Files</pre> folder we have folders with names like <pre>030605.A.b02</pre>. What does the <pre>b02</pre> part mean? Is it the code for the test subject (cat)? Or it is an ID of a neuron?

3) Inside the folder <pre>030605.A.b02</pre> there are files with names <pre>030605.A.b02cequalpower_B3.sa0</pre>, <pre>030605.A.b02dequalpower_B3.sa0</pre>, <pre>030605.A.b02eequalpower_B3.sa0</pre>. What those letters <b>c, d, e</b> indicate?

4) The description of the data says that the data was recorded from 60 channels. But for every <pre>.sa0</pre> file when I load it using <pre>fget_hdr</pre> the <pre>hdr.DataInfo.Channel</pre> is always 0. What is the correct way to interpret this number?

Ilya, I have posted some answers to your questions below. This data was collected some time ago, so my recollections are limited. However, I have reviewed some of my data and notes and can hopefully shed some light on your questions:

1) I believe this is referring to the three classes of natural image sequences (Equalpower A/B/C). The differences between these classes are minimal but worth noting. The A sequences have lower RMS contrast. This was done to minimize the inevitable [0 255] clipping that can occur when image patches are re-scaled. The B sequences have higher RMS contrast and thus more clipping (I don't remember the fraction of pixels that were clipped, but it relatively small < 5% of pixels). The C sequences have the same contrast as B sequences with both uniform (i.e. spatial frequency = 0) and redundant frames (i.e. correlation coefficient between existing frames > 0.95) removed. This helps when inverting the stimulus auto-correlation matrix. The numbers simply indicate the order in which they were created.

2-3) File naming convention:

YYMMDD.(experimental chamber [A,B]).(electrode penetration indicator [a-z])(cell number [00-99])(run number [a-z])(stimulus name, e.g. 'equalpower_B1').(file type, e.g. 'sa0')

030605.A.b02cequalpower_B3.sa0 = Spike file collected June 5, 2003; Chamber A; Penetration b, cell number 2, run number 3, using equalpower_B3 stimulus.

4) I'm not 100% sure of this, but basically the recording system could be configured for up to 60 channels of simultaneous recording. However, for this dataset we were primarily using one electrode (sa0). Typically, there was only one cells that reached threshold at a time/depth, however, occasionally there were multiple cells that we separated by clustering methods (sr0, sr2, ...).

Hope this helps.

- Jon Touryan

Previously Ilya Kuzovkin wrote:

Hi community

My name is Ilya, I am a PhD student at the Computational Neuroscience lab in University of Tartu and I have a few questions about the PVC-2 dataset.

I am interested in the part of the dataset related to the natural images (<pre>crcns-pvc2/2D_noise_natural</pre> and I have a trouble understanding the structure of the data​:

1) In the articles it is mentioned that "three distinct natural ensembles were used in this study", but there are 8 <pre>Equalpower*</pre> files inside the <pre>Stimulus_Files</pre>. Why is that? What letters A,B,C and numbers 1,2,3 mean? Are these different sequences of images or same images but in different order?

2) Inside the <pre>Spike_and_Log_Files</pre> folder we have folders with names like <pre>030605.A.b02</pre>. What does the <pre>b02</pre> part mean? Is it the code for the test subject (cat)? Or it is an ID of a neuron?

3) Inside the folder <pre>030605.A.b02</pre> there are files with names <pre>030605.A.b02cequalpower_B3.sa0</pre>, <pre>030605.A.b02dequalpower_B3.sa0</pre>, <pre>030605.A.b02eequalpower_B3.sa0</pre>. What those letters <b>c, d, e</b> indicate?

4) The description of the data says that the data was recorded from 60 channels. But for every <pre>.sa0</pre> file when I load it using <pre>fget_hdr</pre> the <pre>hdr.DataInfo.Channel</pre> is always 0. What is the correct way to interpret this number?

 

Hi!

Thanks for sharing the data. I know it has been a long time!

I would also like to know if data is collected from a different animal every recording day (so about 60 animals), or if for some days recordings came from the same animal.

Thanks

- Richard Naud