Perceptron & Neural Network project

Python university project for implementing a basic perceptron and a neural network for solving several problems.

Requirements

Several libraries are used. Mainly numpy, pandas and sklearn(shuffle util). If when running any dependency issues arise, PIP tool should be used for solving such issues.
Implemented and tested on Python 3.7

Running

Navigate to bundles root dir. Next run each of the requested exercises. Each file name in this directory indicates the exercise.
All files can be configured inside. And run with python file. Pyton path must be set like PYTHONPATH=src. @see section on example runs

Configurations

• learning_rate => % of delta W to add to weighs on neural network
• momentum => momentum variable. Ideally in [0,1]
• iteration_limit => number of epochs
• features => number of input arguments
• activation_function => the function to be used from src/ar/edu/itba/sia/group3/Functions/Activations_Functions.py.
• restart_condition => used for restarting perceptron on error not changing
• layer_info_list => defines the neural network structure.

More on layer_info_list

Firtst element is first hidden layer, the following are the remaining hidden layers in order. Final element is the output. All elements have 2 int parameters. First int is number of neurons in layer, second int is number of neurons or inputs in previous layer.

Example runs

From root directory.
PYTHONPATH=src python3.7 src/ar/edu/itba/sia/group3/ejer1_and.py
PYTHONPATH=src python3.7 src/ar/edu/itba/sia/group3/ejer2.py