Discover hongyusu/RTA Open Source project by Aalto (@hongyusu)

Stars
1
Language
MATLAB
License
GNU General Publi...
Created almost 11 years ago
Updated almost 10 years ago

hongyusu

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

The repository contains codes and some experimental data for Random Tree Approximation (RTA) algorithm, which appears in Advances in Neural Information Processing System (NIPS 2014). The algorithm is designed for multilabel classification when the output graph is not known apriori. It performs Max-margin Learning on a collection of random spanning trees sample from a complete graph connecting multiple labels. RTA also lays the foundation of performing inference on any unknown graph structures.

SGNMF

NMF with graph laplacian and sparsity constraints

MATLAB

bigdata_etl

Java

sentiment_predictor

http://sentipred.heroku.com

JavaScript

L2RTA

http://hongyusu.github.io/L2RTA This repository contains RTA algorithm with improved inference algorithm for update direction finding. The RTA algorithm was originally introduced in Advances in Neural Informatics Processing System NIPS 2014. In each stochastic gradient descent iteration, the new inference algorithm aims to combine multiple updated directions. In other words, the optimal updated direction is given by a convex combination of many suboptimal update directions. Thus, the new inference algorithm avoid computing the best direction from the K-best list which was used in original RTA algorithm.

MATLAB

mmcrfmissing_scripts