SAUCIE_PyTorch

PyTorch Implementation of SAUCIE Algorithm

For educational (or just enjoying) purposes only. Not fully tested for real dataset.

Requirements

numpy 1.16.2
pandas 0.24.1
pytorch 1.3.0

Difference from the original implementation.

• Tensorflow old version => PyTorch
• Bug? fixes.
• Adjusted the impractical parameters (e.g. sigmas in RBF kernel function).
• Added the resolution parameter in the cluster merging process (merge_k_nearest).
• Modified Maximum Mean Discrepancy calculation. Multiple batch corrections can be learned simultaneously.

Usage

See notebook