Human ochr ensembl regulatory dataset

[simecek]

Check distribution length & train basic CNN (to be sure it can be predicted)

[katarinagresova]

CNN is unable to learn. Will not use this dataset.

[katarinagresova]

There might be a problem with PyTorch code.

[katarinagresova]

Sorry for so many changed files. I needed to rebase main to get code for experiments. Files important for this review are:

• https://github.com/ML-Bioinfo-CEITEC/genomic_benchmarks/blob/human_ochr_ensembl_regulatory/docs/demo_human_ocr_ensembl_regulatory/README.md
• https://github.com/ML-Bioinfo-CEITEC/genomic_benchmarks/blob/human_ochr_ensembl_regulatory/docs/demo_human_ocr_ensembl_regulatory/create_datasets.ipynb
• https://github.com/ML-Bioinfo-CEITEC/genomic_benchmarks/blob/human_ochr_ensembl_regulatory/experiments/tf_cnn_experiments/demo_human_ocr_ensembl_regulatory.ipynb

I was not able to make PyTorch CNN learn something from this data, but TF got accuracy 73%.