MDFT_NN

A recurrent neural network for learning MDFT parameters.

Install package

Run the following line for installation

pip install git+https://github.com/Rahgooy/MDFT.git@master

Reproducing the results of my thesis

Data generation

The generated data are available in the /data folder. However if you want to regenerate them use the following instructions. The data generation method is written in Matlab.

• Run ./matlab/BuildsimMultiMDF.m to build the mex files if you are not using a mac system.

• Run ./matlab/data_generator.m file to generate the data.

Maximum Likelihood (MLE) results

The results are saved in the ./results/MLE folder. If you want to run the model run ./matlab/learn_multi.m file.

Neural Net (NN) results

The results are saved in the ./results/NN folder. If you want to run the NN model in the root folder run bash ./scripts/learn.sh.

Summary of the results

To see the summary of the results saved in the ./results/MLE and ./results/NN folders run:

python summarize.py