Reviews

[anhhuyalex]

python figs.py alex_aug_24_interleave --data_dir=/scratch/qanguyen/schlichting --searchvar=LRateOverAll
python figs.py alex_aug_24_blocked --data_dir=/scratch/qanguyen/schlichting --searchvar=LRateOverAll

and same day with favila

TO DO: rerun job with more parameters, just use boundary_condition_plots ipynb script

[anhhuyalex]

python figs.py alex_aug_29_different --data_dir=/scratch/qanguyen/favila --searchvar=LRateOverAll

[anhhuyalex]

For instance, it's clear from the manipulation of oscillation strength in the model of Schlichting et al that this parameter can dramatically change the direction of the results. The authors do report the oscillation strength parameter values that they used in the other two models, but it is not clear how sensitive these models are to small changes in this value.

ACTION ITEM
Since both reviewers 1 and 2 seem interested in understanding the consequences of shifting osc_amount, we can run a new simulation with a range of osc_amount, then decide next steps (whether to include whole simulation w results, or just give a sentence)

I’m not sure that R1 is interested in this. But if it’s easy enough, we could try manipulating osc amount in the Chanales and Favila sims to see what happens (e.g., it’s possible that slightly increasing osc size in favila will lead to a situation where you get some differentiation in the different face condition but more in the same face condition, which aligns better with the actual data – see reviewer comment towards the end of this doc)

Similarly, it's not clear whether the 2/6 hidden layer overlap (only explicitly manipulated in the model of Chanales et al) is required for the other two models to work.

ACTION ITEM:
Clarify in favila intro WHY we chose the 2/6 overlap as the model.
(shorter version of the following): We chose the 2/6 version as the framework for the other two studies because we built the model of Chanales such that differentiation would happen in 2/6, and for consistency it was helpful to have the same framework across models to show differentiation. We therefore selected parameters such that the 2/6 condition would show differentiation in the other two models. In theory a different level of overlap could lead to differentiation, but since differentiation is highly sensitive to activation dynamics, all the other parameters would have to be adjusted to account for it.

I agree with this plan [also, note that, in practice, i think we may be able to “shift” differentiation to the 3/6 condition by using a different oscillation size – we should try to do this]

[anhhuyalex]

python figs.py alex_sep_5_oscamnt --data_dir=/scratch/qanguyen/chanales --searchvar=Hidden_ColorRecall_Layer_OscAmnt
python figs.py alex_sep_5_oscamnt --data_dir=/scratch/qanguyen/favila--searchvar=Hidden_ColorRecall_Layer_OscAmnt