HTN

Hidden Tree Markov Network

The official Python implementation for the Hidden Tree Markov Network (HTN), a mixed neural-generative model for learning with tree structured data. The HTN allows to integrate features adaptively learned by multiple probabilistic hidden tree Markov models (possibly leveraging different parsing directions) using a neural-based fusion layer and end-to-end training by backpropagation.

The code is maintained in the Github of my student Valerio De Caro, who is to be credited for the implementation. The code exploits Tensorflow (TF2.0) support and optimizations and can be downloaded here.

The code is provided as is with no warranty and technical support. Please inform the authors of the original paper (details below) and of the code if you intend to redistribute the code.

Citation

If you find this code useful, please remember to cite:

Bacciu Davide: Hidden Tree Markov Networks: Deep and Wide Learning for Structured Data. Proc. of the 2017 IEEE Symposium Series on Computational Intelligence (SSCI'17), IEEE, 2017.

BibTeX (Download)

@conference{dl2017,
title = {Hidden Tree Markov Networks: Deep and Wide Learning for Structured Data},
author = {Bacciu Davide},
url = {https://arxiv.org/abs/1711.07784},
year  = {2017},
date = {2017-11-27},
booktitle = {Proc. of the 2017 IEEE Symposium Series on Computational Intelligence (SSCI'17)},
publisher = {IEEE},
keywords = {deep learning, hidden tree Markov model, structured data processing},
pubstate = {published},
tppubtype = {conference}
}