A personal Machine Learning MATLAB toolbox. Algorithms are implemented in a simple and readable way.
Part of this code is inspired by the code of Coursera's Machine Learning course by Andrew Ng. If you arrived here looking for a Deep Learning toolbox you probably were looking for Lasagne, only related to this in a gastronomical way.
The main use of this toolbox is education and research and, according to its name, I do not recomend it for data-intensive production environments. In such a case, you can contact me at [email protected] for profesional advice.
Instalation:
Run besaml_setup.m
file before using the toolbox.
Current Version:
- Modelling of data using a Gaussian Mixture Model (GMM) fitted using Expectation-Maximization (EM).
- Multiclass Softmax Regression Classifier.
Dataset included:
I have included some small datasets to test the implemented algorithms.
- Old Faithful Geiser Dataset [1,2].
- Subset of MNIST dataset [3] selected in Coursera's Machine Learning course.
Function fmincg
. Copyright (C) 2001 and 2002 by Carl Edward Rasmussen. Date 2002-02-13.
Development:
Besaml is a work in progress, any comment would be welcomed.
References:
[1] Hardle, W. (1991) Smoothing Techniques with Implementation in S.
New York: Springer.
[2] Azzalini, A. and Bowman, A. W. (1990). A look at some data on
the Old Faithful geyser. Applied Statistics 39, 357-365.
[3] Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner. "Gradient-based
learning applied to document recognition." Proceedings of the IEEE, 86(11):2278-2324, November 1998.