Some of the experiments we are currently tracking (.csv files with Git, and experiments outputs with DVC) are not needed any more, and they will not be included in the paper.
We should clean up the repository, so that also a dvc repro would be much faster, and we can run benchmarks and perform an analysis of the results.
Actions
Remove unnecessary files for the following experiments.
[x] data/final/IN_data_no_NBC.csv
subset of data/final/IN_data.csv: the difference is that data/final/IN_data_no_NBC.csv does not include cells with M-type NBC
used by features-lida-in-merged-no-nbc-*** stages
outputs written to lida-in-merged-no-nbc
outputs are used by train-lida-in-merged-no-nbc stage
checkpoints written to checkpoints-lida/in-merged-no-nbc-***
@lidakanari said [2022-05-03] : this was just an experiment, not necessary to keep any more, we are not going to use it in the paper
[x] data/final/interneurons-merged-layers.csv
an alternative to data/final/IN_data.csv, created by @Stannislav while waiting for data/final/IN_data.csv to be provided by @lidakanari
subset of data/final/IN_data.csv: the difference is that data/final/interneurons-merged-layers.csv does not include samples with M-type BTC (found only in L23, L4, and L5) and CHC (found only in L23 and L5)
used by features-in-merged-layers-*** stages
outputs written to in-merged-layers
outputs are used by train-merged-interneurons
checkpoints written to checkpoints-merged-interneurons/***
@lidakanari said [2022-05-03] : this seems an old, deprecated version of data/final/IN_data.csv, so we can remove
[x] data/final/pc-L2-merged.csv
takes neurons of pc from layer L2 and merges M-types: all the the samples with M-type TPC_X (i.e. TPC_A, TPC_B, TPC_C) are here just given the shared M-type TPC
used by features-pc-L2-merged-*** stages
outputs written to pc-L2-merged
outputs are used by train-pc-L2-merged-classes
checkpoints written to checkpoints-pc-merged-classes/pc-L2-***
@lidakanari said [2022-05-03]: probably it will not be used for the paper, we can remove
Context
Some of the experiments we are currently tracking (
.csvfiles with Git, and experiments outputs with DVC) are not needed any more, and they will not be included in the paper.We should clean up the repository, so that also a
dvc reprowould be much faster, and we can run benchmarks and perform an analysis of the results.Actions
Remove unnecessary files for the following experiments.
data/final/IN_data_no_NBC.csvdata/final/IN_data.csv: the difference is thatdata/final/IN_data_no_NBC.csvdoes not include cells with M-typeNBCfeatures-lida-in-merged-no-nbc-***stageslida-in-merged-no-nbctrain-lida-in-merged-no-nbcstagecheckpoints-lida/in-merged-no-nbc-***data/final/interneurons-merged-layers.csvdata/final/IN_data.csv, created by @Stannislav while waiting fordata/final/IN_data.csvto be provided by @lidakanaridata/final/IN_data.csv: the difference is thatdata/final/interneurons-merged-layers.csvdoes not include samples with M-typeBTC(found only inL23,L4, andL5) andCHC(found only inL23andL5)features-in-merged-layers-***stagesin-merged-layerstrain-merged-interneuronscheckpoints-merged-interneurons/***data/final/IN_data.csv, so we can removedata/final/pc-L2-merged.csvpcfrom layerL2and merges M-types: all the the samples with M-typeTPC_X(i.e.TPC_A,TPC_B,TPC_C) are here just given the shared M-typeTPCfeatures-pc-L2-merged-***stagespc-L2-mergedtrain-pc-L2-merged-classescheckpoints-pc-merged-classes/pc-L2-***data/final/pc-L5-merged.csvdata/final/pc-L2-merged.csvdata/final/pc-L6-merged.csvdata/final/pc-L2-merged.csv