CSC591/ECE 542 - Neural Networks: Proj-C

Team 116

• Ameya Vaichalkar (agvaicha)
• Keerthana Telaprolu (ktelapr)
• Vikram Pande (vspande)

This readme contains the directory structure. The code has modular structure and there are files for different functionalities. Final predictions are in the jupyter notebook file.

Following are the modules we built for the project:

• helper_functions.py: contains the functions to plot the graphs of accuracy and loss against epochs, get the evaluation metrics and function to assign weights to imbalanced classes.
• lstm_models.py: functions of variants of LSTM models with different combinations of parameters such as LSTM, BiDirectional LSTM, Dropout, Learning Rate.
• notebook.ipynb: contains the code for predictions with modular methods.
• preprocess.py: contains methods to preprocess the data. Methods to normalize the data, encode the data, match frequency of X and Y, get windowed data to feed LSTM, get the data in time series format.
• templates.py: contains templates of train and test datafiles and some constants

How to run? Steps for running the code:

1. The directory contains all the necessary files, download/clone the repository.
2. Copy the data in the same directory.
3. Run notebook.ipynb that generates the prediction files.

References:

[1] B. Zhong, R. L. d. Silva, M. Li, H. Huang and E. Lobaton, "Environmental Context Prediction for Lower Limb Prostheses With Uncertainty Quantification," in IEEE Transactions on Automation Science and Engineering, vol. 18, no. 2, pp. 458-470, April 2021, doi: 10.1109/TASE.2020.2993399.