Data set for 2P for NMA CN project (image categories)

HI,
I was participating in project for Neuromatch academy computational neuroscience. I was working on the data in the Allen data set at Google Colaboratory.

When we look at the data from allen_visual_behavior_2p_change_detection_familiar_novel_image_sets:

There are two confusion:

there seems to be mismatch among novel-familiar exposure_level column and the session_type column. For examle “OPHYS_3_images_B image” are in the same row as is the “familiar”. And “OPHYS_4_images_A” are in the same row as is the “novel”. it seems a mismatch since set “A” images are presented as “familiar” and sets “B, C, D” images are presented as “novel” in the behaviour paradigm.

we are not able to see data for images of “C” and “D” set (with .unique()) along any cell cre_line. Is it there or we are not looking correctly.

Although in the paper https://elifesciences.org/articles/55130 all the image sets are used: A, B, C, D.

Also, are the B, C and D sets are avialable in the extended data set: Allen Visual Behavior from SDK (in the end the tSNE plot has images which correspond to the image number (im000, im031… to im106) of set C (not all sets) discussed in the paper. If they are there, are they categorized as sets B, C, D or we will have to find them by image number? Thanks!

Could you please guide us through.

Hi,

Your question about the relationship between familiar/novel, exposure level, and session_type should be addressed by the response to another forum post here.

Another important thing to note is that the eLife paper you linked did not use the same dataset as the Visual Behavior 2P project dataset you are using as part of the Neuromatch course. The eLife paper used a pilot dataset that we collected prior to launching large-scale data collection for the Visual Behavior project.

We used these pilot experiments to get initial results and refine our experimental design for the larger dataset. For example, in the Visual Behavior dataset, we only use 1 novel image set, but we include multiple sessions with that image set to allow tracking of changes in neural activity as novel images becoming increasingly familiar across multiple exposures, whereas the eLife paper only includes one session for each novel image set.

So you are correct that image sets C and D are not available in the dataset you are working with.

Please let me know if you have additional questions.

Thanks a lot! that clears a lot of confusion.