multi class classificatio

[vinayakumarr]

I have large amount of sequence data
20 features and 5 classes for training
20 feature and 5 classes for testing

Can you tell whether it is possible to apply your method for classifying this? if so could you please give a simple example

[cytsai]

Hi, thanks for your interest in our project.

Since our code is still quite experimental and research-oriented now, I would actually recommend trying other more popular libraries (e.g. scikit-learn, Vowpal Wabbit, XGBoost, etc.) first.

But if you don't mind spending some time tweaking our code, here are some tips, which hopefully can help a bit. The main thing needs to be modified is the load function in tsnet/datasets.py, where you should shape X_trn and X_tst as Nx20x1x1 arrays (N being the number of data points), and y_trn and y_tst as Nx1 arrays (with entries valued from 0 to 4 in your case). Then you can follow our MLP examples to create your own networks (nips16/experiments/cmp_mlp.py).