jazz_melody_generation

This repository provides the following methods to reproduce the results for our APSIPA paper:

• VAE by multi-task training
• VAE by adaptation training
• Objective evaluation

Requirements

The following requirement is not a minimum set but should cover packages needed to run everything.

• torch
• subprocess
• time
• torchvision
• matplotlib
• numpy
• mgeval

Basic Usage

There are two part of codes.

1. The architecture and the recipe for training in the paper.
2. The Evaluation code.

TODO

• architecture and the recipe for training in the paper
• The Evaluation code

Papers and Audio samples

• APSIPA 2019 conference paper: [arXiv] [Demo]

Authors

• Hsiao-Tzu Hung, Institute of Information Science, Academia Sinica, Taipei, Taiwan. Taiwan AI Labs, Taipei, Taiwan.
• Chung-Yang Wang, Taiwan AI Labs, Taipei, Taiwan.
• Yi-Hsuan Yang, Research Center for IT Innovation, Academia Sinica, Taipei, Taiwan. Taiwan AI Labs, Taipei, Taiwan
• Hsin-Min Wang, Institute of Information Science, Academia Sinica, Taipei, Taiwan.