Skip to content

Latest commit

 

History

History
57 lines (45 loc) · 2.57 KB

README.md

File metadata and controls

57 lines (45 loc) · 2.57 KB
Raw

Multi-layer Perceptron Model

This project implements a feed-forward multi-layer perceptron model.

It contains two main classes for training and running.

Class Name Purpose
Main Training and Exporting Networks
RunNetwork Importing and Running Networks

General Usage

For neatness, all log files will be placed in the logs folder. While all valid file paths are supported, please export networks to the networks folder.

Compilation

To compile this project, navigate into the project and execute javac *.java.

Training and Exporting Networks

To run, execute java Main. Enter the requested values. Only networks with one hidden layer will train.

To compile, run, and delete all class files afterwards, run javac *.java && java Main && rm *.class

Reads inputs from console. Sample training inputs can be found in the trainingCases folder.

In these files, all values are separated by spaces or carriage returns. These files are formatted as follows.

  • Network
    • The number of input nodes (integer)
    • The number of hidden layers (integer)
    • The number of nodes in each hidden layer (integers)
      • For 3 hidden layers with 2, 4, and 6 nodes (from the input side to output side), this would be 2 4 6
    • The number of output nodes (integer)
    • The range in which weights will be randomized (two decimals)
  • Training Cases
    • The number of training cases (integer)
    • For each case:
      • The case's inputs (decimals)
      • The case's outputs (decimals)
  • Training Parameters
    • The initial training factor, lambda (decimal)
    • The adaptive training constant (decimal)
    • The maximum number of steps (integer)
    • The ceiling of the desired error (decimal)
    • The lower limit of the training factor (decimal)
    • The period of steps at which updates are printed (integer)
      • If set = 0, no updates will occur
    • The period of steps at which the network is saved (integer)
      • If set = 0, no saves will occur
  • Exporting
    • The file where the network should be exported to. Usually networks/DESCRIPTIVE_NAME.txt.

Importing and Running Networks

After running training, networks will be exported to files. Run these files with the java RunNetwork.

To compile, run, and delete all class files afterwards, run javac *.java && java RunNetwork && rm *.class

To submit to Dr. Nelson:

Run rm -f *.zip && rm -f *.class && zip -u -r bennett_liu_perceptron.zip * && zip -d -r bennett_liu_perceptron.zip networks/* to create a zip of only Java code, folders, and README.md