MUPET: Sparse subspace clustering

This repo contains code to cluster mouse vocalizations using sparse subspace clustering.

Instructions

1. Download a sample dataset from here.

2. Preprocess the data

   • Use MUPET to create a dataset
   • Load workplace/datasets/~.mat and run prepare_data.m to collect the syllables detected by MUPET
   • A file such as C57_all_RUs.mat is generated
   • Note: for the DBA dataset, we only use the first 9000 syllables
   • Run segment_whole_dataset.m to separate the whole dataset into inlier dataset and outlier dataset

3. Clustering: you can use jupyter notebook "main" to get the results step by step

4. Plot centers and cluster outliers: run look_center.m to depict the cluster centers