Dear Supriyo Srimani ,

Please, take a look at my matlab toolbox on ensemble learning: https://www.mathworks.com/matlabcentral/fileexchange/68383-ensemble-learning-toolbox

It works well with matlab's statistical and machine learning toolbox.

Unfortunately, matlab's neural network toolbox is not fully compatible, so you'll need to embed it with custom functions. (Feel free to contact me if you need help with this)

Dear Victor Henrique Alves Ribeiro,

Thanks for the toolbox. It is extremely helpful.

I need help regarding the ensemble of neural network with the existing toolbox of yours. The neural network training mus be added at "prepare learner section". I have attached a neural network code using neural network toolbox but I am a novice in writing the custom function (as you have suggested). Can you help me with this.

Dear Supriyo Srimani ,

I have prepared a custom class to embedd the patternnet.

I have attached such a class and also a new example based on neural networks ensembles. (there is a commented line that shows how to train NN + SVM ensembles)

I hope this helps you in your research!

Please, let me know if this works for you.

Dear Victor Henrique Alves Ribeiro

Thank you for your help in developing the code.

I have tested the code on wine data set by using SVM and NN. the 'fitcecoc' function for SVM does not give good accuracy whether I have developed the custom code for multiclass SVM which gives a good result. Can you help to integrate this custom code for SVM with your classifiers? You can include this to your toolbox also if you feel.

I am attaching the folder in which all your functions along with my functions are embedded.

1. Wine Test is the code using fitcecoc and fitcnet

2. wine svm is the custom code for multiclass svm

Can you please help me regarding this

Dear Supriyo Srimani ,

I am sorry, but I do not have the Bioinformatics toolbox to run part of the experiment.

However, I was able to achieve good results with SVM in the wine dataset (similar to NN, at least).

The only modification in wine_test.m is to set 'Standardize' to 1 when creating the templateSVM object.

gaussian = templateSVM('KernelFunction', 'gaussian', 'Standardize', 1);

I hope this can help you in your research, as I am currently not able to use the bioinformatics to test your implementation of multiclass svm.