This paper aims to create two deep learning networks, one from scratch where we refrain from state-of-art neural network packages and instead we use basic python libraries, and other using machine learning framework, in particular Keras frame- work. We provide a thorough exploratory data analysis and base the decision in our neural networks outcomes. Both networks are created dynamically in order to evaluate different configurations and identify the optimal number of layers and perceptrons. Both model metrics are compared and next steps are provided to improve the performance.
This project creates two different NN to binary classify the given dataset
See the paper of this project for a detailed information about each NN defined. We also refer the reader to the paper for specific details about the NN.
Use the following commands to run the NN
- pip install -r requirements.txt
- python3 main.py
Control parameters are in the main.py file
The run instructions were tested in MacOS using Python 3.8.8
Please note that running the entire script can take up to 3 minutes depending on the performance of the computer that it is ran on.