FedscGen: PyTorch randomly initialized models

The provided files correspond to models trained using PyTorch for batch effect correction and cell type annotation across various datasets. These models have been employed in the FedscGen framework, a federated learning approach designed for handling batch effect corrections in multi-dataset scenarios. Meanwhile, models in `classification` zip file have been utilized for training classifiers to annotate cell types.

Directory Structure and File Descriptions:

• CellLine, HumanDendriticCells, HumanPancreas, MouseBrain, MouseCellAtlas, MouseHematopoieticStemProgenitorCells, MouseRetina, PBMC:
  • These zip files contain models and related metadata specific to each dataset.
  • Files in Each Dataset Directory
    • attr.pkl: A pickle file containing attributes related to the dataset and model, such as configuration details and training metadata.
    • model_params.pt: A file storing the model parameters of the PyTorch model, which were used for training on the respective dataset.
    • var_names.csv: A CSV file listing variable names or features used in the dataset.
• classification:

• • CellLine, HumanDendriticCells, HumanPancreas, MouseBrain, MouseCellAtlas, MouseHematopoieticStemProgenitorCells, MouseRetina, PBMC:
    • Each subdirectory contains a model.pth file, which is the PyTorch model checkpoint for the classifier trained to annotate cell types for each respective dataset.