Information Maximizing Self Augmented Training (IMSAT)
This is a reproducing code for IMSAT . IMSAT is a method for discrete representation learning using deep neural networks. It can be applied to clustering and hash learning to achieve the state-of-the-art results. This is the work performed while Weihua Hu was interning at Preferred Networks.
You must have the following already installed on your system.
- Python 2.7
- Chainer 1.21.0, sklearn, munkres
For reproducing the experiments on MNIST datasets in , run the following codes.
- Clustering with MNIST:
- Hash learning with MNIST:
calculate_distance.py can be used to calculate the perturbation range for Virtual Adversarial Training . For MNIST dataset, we have already calculated the range.
 Weihua Hu, Takeru Miyato, Seiya Tokui, Eiichi Matsumoto and Masashi Sugiyama. Learning Discrete Representations via Information Maximizing Self-Augmented Training. In ICML, 2017. Available at http://arxiv.org/abs/1702.08720
 Takeru Miyato, Shin-ichi Maeda, Masanori Koyama, Ken Nakae, and Shin Ishii. Distributional smoothing with virtual adversarial training. In ICLR, 2016.