nickledave / regularizeddctrainingfordmn

Regularized divide and conquer training for dendrite morphological neurons

Regularized divide and conquer training for dendrite morphological neurons

This package contains functions to train dendrite morphological neurons based on the paper "Regularized Divide and Conquer Training for Dendrite Morphological Neurons", E. Zamora, H. Sossa, 2016. This method is effective to overcome overfitting in classification problems for machine learning.

[Please cite it if you intend to use this code]

Installation and Demo

It is not needed. Just download files where you want and run "demo_for_training" on Matlab R2015a or later.

[Coming soon...a video for a briefly explanation of the training algorithm YouTube]

Experiments

You can reproduce the experimental results of the paper by running "main_DatasetsTraining.m" to reproduce experimental results in each dataset. The files "RegularizedCurves_for_E0.m" and "RegularizedCurves_for_L0.m" generate the regularized curves for tolerated error E0 and depth limit L0, respectively.

Please see the code for further details.

Acknowledgments

The authors would like to acknowledge the support for this research work to the UPIITA and the CIC, both schools belong to the National Polytechnic Institute (IPN). This work was economically supported by SIP-IPN under grants 20160945 and 20161116, and CONACYT, under grant 65.