Label flip genomic benchmark dataset

#1
by isabelku - opened

In the linked dataset on which the model is fine-tuned, the labels are flipped compared to the official GitHub repository of the genomic benchmark (https://github.com/ML-Bioinfo-CEITEC/genomic_benchmarks/tree/main/datasets/human_enhancers_cohn). The same sequence in the linked dataset here has the label 0 when on the official git it is 1, and vice versa. I wanted to ask if this flip is intentional and, if so, what is the reason for it?

Sign up or log in to comment