Simple and Multiple Layer Perceptron Implementation

This project aims to implement a Simple and Multiple Layer Perceptron using Python 3.10. It includes a pipfile that lists all the dependencies required to run the project.

Dependencies

• Python 3.10

Usage

There are five run scripts in this project, each of which receives a configuration file as an argument:

• run_1.py - runs Simple Perceptron with ex1.json configuration file.
• run_2.py - runs Simple Perceptron with ex2.json configuration file.
• run_3a.py - runs Multiple Layer Perceptron with ex3a.json configuration file.
• run_3b.py - runs Multiple Layer Perceptron with ex3b.json configuration file.
• run_3c.py - runs Multiple Layer Perceptron with ex3c.json configuration file.

| Each script corresponds to a specific challenge

To run any of the scripts, first install the dependencies using pipenv:

pipenv install

Then run the desired script with the appropriate configuration file:

pipenv run python run_X.py exX.json

Configuration files

Each configuration file is in JSON format and specifies the parameters for the corresponding perceptron. Here is an example of the structure of the configuration file:

{
  "activation_function": {
    "function": "sigmoid",
    "beta": 0.8
  },
  "cut_condition": {
    "condition": "mse",
    "eps": 0.001
  },
  "optimization_method": {
    "method": "gradient",
    "learning_rate": 0.1
  },
  "epochs": 500
}

• activation_function - an object that specifies the activation function used by the perceptron. It contains:
  • function - a string representing the name of the activation function (e.g., "sigmoid", etc.)
  • beta - a float representing a parameter of the activation function (e.g., the slope of the sigmoid function)
• cut_condition - an object that specifies the stopping criterion for training the perceptron. It contains:
  • condition - a string representing the name of the stopping criterion (e.g., "mse" for mean squared error)
  • eps - a float representing the threshold value for the stopping criterion
• optimization_method - an object that specifies the optimization method used for training the perceptron. It contains:
  • method - a string representing the name of the optimization method (e.g., "gradient" for gradient descent)
  • learning_rate - a float representing the learning rate used by the optimization method
• epochs - an integer representing the number of epochs to train the perceptron for