ntienvu/NonparametricBudgetedSGD

Matlab code for Nonparametric Budgeted SGD for classification and regression (AISTATS 2016)

Description

This is the released source code for Nonparametric Stochastic Gradient Descent.

Run the code

run demo_NBSGD_binary_classification for binary classification
run demo_NBSGD_multi_classification for multiclassification
run demo_NBSGD_binary_classification for regresion

Explanation of some parameters

There are three main parameters. lambda: for regularization. The default value is 100/N where N is the number of training data points. sigma: for RBF kernel. The defaul value is 0.1 beta: control the budget maintenance rate, ranging from 0 to 1. The default value is 0.5

Contact: Dr Vu Nguyen, vu@ieee.org

Citation:

Le, T., Nguyen, V., Nguyen, T. D., & Phung, D. (2016). Nonparametric Budgeted Stochastic Gradient Descent. In Proceedings of the 19th International Conference on Artificial Intelligence and Statistics (pp. 654-572).