dance
dance copied to clipboard
OmicsML
DANCE: a deep learning library and benchmark platform for single-cell analysis
Use [omegaconf](https://github.com/omry/omegaconf) `DictConfig` object for storing and managing configurations. Update preprocessing pipelines construction with config parsing. 1. Data selection 1. Preprocessing pipelines 1. Model params and pipelines 1. Eval pipelines...
TODO in a future PR: refactor dataset object to use the dance abstract dataset class, see [`ClusteringDataset`](https://github.com/OmicsML/dance/blob/f3e0e83ef3dbf25339e1a6ffa10ffb18ac81d72d/dance/datasets/singlemodality.py#L235) for example _Originally posted by @RemyLau in https://github.com/OmicsML/dance/pull/228#discussion_r1129447102_
- [ ] Use MSE - [ ] Compute on `log1p` normalized expression instead of the raw counts
Currently, there are several different dataset objects specialized for each task and model (e.g., [`CellTypeDataset`](https://github.com/OmicsML/dance/blob/79de295feadc5219e6e5ed9e8b95f7b3c94042fc/dance/datasets/singlemodality.py#L44), [`ClusteringDataset`](https://github.com/OmicsML/dance/blob/79de295feadc5219e6e5ed9e8b95f7b3c94042fc/dance/datasets/singlemodality.py#L494)), each of them takes a variety of specialized arguments that are not directly related...
- [ ] TODO later: use the `download_unzip` utility instead of doing these system calls. _Originally posted by @RemyLau in https://github.com/OmicsML/dance/pull/223#discussion_r1120031057_
Some current graph `transforms` methods directly save the processed `dgl` graphs in `.uns`. The limitations for doing so are - Limits the choice of the framework (`dgl` vs. `pyg`) to...
We need both the return types for `feature_channels` and `label_channels`..
- [ ] `index` option - [ ] `mask` option