/comp_bio_ex3

Primary LanguagePython

Ex3 - Neural Network with Genetic Algorithm

This project implements a neural network with a genetic algorithm approach to learn patterns in binary strings. The goal is to build a network that can predict the legality of new binary strings based on the learned patterns. There are two separate problems addressed in this project, each requiring a different program.

Programs

Buildnet0

Buildnet0 program is designed to build a neural network for the first legality problem. It takes a learning file and a test file as input and outputs the structure definitions of the network and the weights of the network.

To run buildnet0 program:

  1. Clone the repository: git clone https://github.com/alonravid1/comp_bio_ex3.git
  2. Two options to set the data:
    • run by defalut: Place the training and test data files in the same folder as the buildnet0 program. Name the files training_set0.txt and test_set0.txt. When the program asked the file paths press enter.
    • run by giving the train and test data path files - the program will ask for path of each, write the path and press enter.
  3. Double-click on the buildnet0.exe file.
  4. When prompted, choose whether to see information about the program (y/n).
  5. Enter the path to the training data file or press Enter to use the default file name.
  6. Enter the path to the test data file or press Enter to use the default file name.
  7. Wait for the program to run. The running time will be displayed.
  8. Press Enter to exit the program.

The output file wnet0.txt will be generated in the same folder as the buildnet0 program.

Runnet0

Runnet0 program takes the output file wnet0.txt generated by buildnet0 and a data file of 20,000 binary strings as input. It predicts the legality of each string based on the learned network and outputs the predictions to a text file.

To run runnet0 program:

  1. Clone the repository: git clone https://github.com/alonravid1/comp_bio_ex3.git
  2. Place the wnet0.txt file generated by buildnet0 and the data file in the same folder as the runnet0 program. Name the data file as desired.
  3. Double-click on the runnet0.exe file.
  4. Press Enter to start the program.
  5. Wait for the program to complete.
  6. The predictions will be saved in the file predictions0.txt in the same folder as the runnet0 program.

Buildnet1

Buildnet1 program is similar to buildnet0 but for the second legality problem. It takes a learning file and a test file as input and outputs the structure definitions of the network and the weights of the network.

To run buildnet1 program:

  1. Clone the repository: git clone https://github.com/alonravid1/comp_bio_ex3.git
  2. Two options to set the data:
    • run by defalut: Place the training and test data files in the same folder as the buildnet1 program. Name the files training_set1.txt and test_set1.txt. When the program asked the file paths press enter.
    • run by giving the train and test data path files - the program will ask for path of each, write the path and press enter.
  3. Double-click on the buildnet1.exe file.
  4. When prompted, choose whether to see information about the program (y/n).
  5. Enter the path to the training data file or press Enter to use the default file name.
  6. Enter the path to the test data file or press Enter to use the default file name.
  7. Wait for the program to run. The running time will be displayed.
  8. Press Enter to exit the program.

The output file wnet1.txt will be generated in the same folder as the buildnet1 program.

Runnet1

Runnet1 program is similar to runnet0 but for the second legality problem. It takes the output file wnet1.txt generated by buildnet1 and a data

file of 20,000 binary strings as input. It predicts the legality of each string based on the learned network and outputs the predictions to a text file.

To run runnet1 program:

  1. Clone the repository: git clone https://github.com/alonravid1/comp_bio_ex3.git
  2. Place the wnet1.txt file generated by buildnet1 and the data file in the same folder as the runnet1 program. Name the data file as desired.
  3. Double-click on the runnet1.exe file.
  4. Press Enter to start the program.
  5. Wait for the program to complete.
  6. The predictions will be saved in the file predictions1.txt in the same folder as the runnet1 program.

Data Format

The input data files (training_set0.txt, test_set0.txt, training_set1.txt, test_set1.txt, and the data file for runnet0/runnet1) should contain one binary string per line. Each string should consist of 16 binary digits (0s and 1s).

Network Structure

The neural network structure used for both problems is defined as [1, 16, 32, 16], where the first number represents the input layer, the last number represents the output layer, and the middle numbers represent hidden layers.

Genetic Algorithm Parameters

The genetic algorithm parameters used for training the networks are specified in the programs and can be adjusted if desired. They include population size, maximum generations, replication rate, mutation rate, learning rate, and tournament size.

Output Files

The output files generated by the programs are as follows:

  • wnet0.txt: Structure definitions and weights of the network trained by buildnet0.
  • predictions0.txt: Predictions of the legality for each string in the data file, generated by runnet0.
  • wnet1.txt: Structure definitions and weights of the network trained by buildnet1.
  • predictions1.txt: Predictions of the legality for each string in the data file, generated by runnet1.