/SoftKMeans

Implementation of Deep Soft-K means

Primary LanguagePython

Overview

Implementation of the Deep Soft-K means algorithm proposed in "Deep clustering: On the link between discriminative models and K-means" availbel at "https://arxiv.org/abs/1810.04246".

Dependencies

Package version used: python 3.5.4 Theano 1.0.1 Lasagne 0.2.dev1

Train model

To run the code for training MNIST-full for instance, you can run $ python soft_K_means.py --dataset "MNIST-full"

Datasets

MNIST-full is uploaded. Other data sets (USPS,Youtube-Face, CMU-PIE, FRGG, ...) may be found in: https://github.com/jwyang/JULE.torch and https://drive.google.com/drive/folders/0B9J-9A2jotGRT25vSDhUWTQxVWs