Basic idea: We want to test how much pretraining on a smaller set of genes helps for increasing performance after finetuning on a larger amount of genes.
This is relevant biologically, since commonly a different set of genes is selected for single cell experiments.
Experiment steps:
Pretrain on LINCS (900 genes), finetune on Trapnell (same 900 genes)
Pretrain on LINCS (900 genes), finetune on Trapnell (2000 genes)
Train from Scratch on Trapnell (900 genes)
Train from Scratch on Trapnell (2000 genes)
Compare performances between CCPA with pretraining and CCPA without pretraining for each of the two settings.
Implementation steps:
[X] @MxMstrmn Generates new Trapnell datasets that have ~2000 genes
[x] @siboehm Writes the transfer code (just adding another layer to decoder & encoder that adjusts the dimensions)
Basic idea: We want to test how much pretraining on a smaller set of genes helps for increasing performance after finetuning on a larger amount of genes.
This is relevant biologically, since commonly a different set of genes is selected for single cell experiments.
Experiment steps:
Implementation steps: