This is a pytorch implementation of the following paper [URL]:
@inproceedings{takahashi2018student,
title={Student-t Variational Autoencoder for Robust Density Estimation.},
author={Takahashi, Hiroshi and Iwata, Tomoharu and Yamanaka, Yuki and Yamada, Masanori and Yagi, Satoshi},
booktitle={IJCAI},
pages={2696--2702},
year={2018}
}
Please read license.txt before reading or using the files.
Please install python>=3.6
, torch
, numpy
and scikit_learn
.
usage: main.py [-h] [--dataset DATASET] [--decoder DECODER]
[--learning_rate LEARNING_RATE] [--seed SEED]
- You can choose the
dataset
from following datasets:SMTP
.- We are preparing other datasets.
- You can choose the
decoder
of the VAE fromnormal
orstudent-t
. - You can also change the random
seed
of the training andlearning_rate
of the optimizer (Adam).
SMTP with Gaussian decoder:
python main.py --dataset SMTP --decoder normal
SMTP with Student-t decoder:
python main.py --dataset SMTP --decoder student-t
- After the training, the mean of log-likelihood for test dataset will be displayed.
- The detailed information of the training and test will be saved in
npy
directory.