Introduction to genetic programming
What is Genetic Programming?
What is Genetic Programming? Part 2
What is genetic programming? Part 3- Symbolic regression
What is genetic programming? Part 4: Modeling chaotic series
What is genetic programming? Part 5: Regime change
Prediction the stock market with genetic programming
Predicting the stock market with genetic programming – Part 1
Predicting the stock market with genetic programming – Part 2
Electronic files to accompany “Implementing the template method pattern in genetic programming for improved time series prediction”, submitted to Genetic Programming and Evolvable Machines
This page describes the files and directories provided in the source code distribution to accompany the article “Implementing the template method pattern in genetic programming for improved time series prediction”, submitted to the journal Genetic Programming and Evolvable Machines.
You can download the binary using the following links
GitHub Source Code Repositories
The provided batch files include one file for each experiment described in the referenced paper. The batches are configured for Microsoft Windows, but are compatible with Linux shell scripts with minor modifications. You must provide the user id and password of your local MySQL/MariaDb installation. The batch files are available in the GitHub repositories.
The batch files are configured to incorporate visualization.The batches can be run on a server by adding the parameter
and setting program parameter
The binary distributions make use of a relational database to store test results. A MariaDB script, adt.sql, is provided in the GitHub repository referenced below . This should also be compatible with MySQL. This script will create a database named adt and the required tables. The included batch files contain username and password parameters which must be changed to work with your local installation.
Coupled vs Decoupled distributions:
The referenced paper describes Automatically Defined Templates (ADT), a new modularity approach for genetic programming. Two approaches to ADT are described Coupled and Decoupled. The initial implementation of ADT assumed a coupled approach. Testing proved this approach was not a reasonable methodology. A change was made to the source code to switch to a coupled approach. Both approaches do not exist as opinions in the same code base. The decoupled approach only exists in earlier versions of the code. Therefore, two source and binary distributions are provided.
The source is distributed in Maven compatible format. The code is distributed under an MIT license.
Two jar files are built from the source described above.
Target nova-1.1-SNAPSHOT.jar Decoupled
The included batch files expect both these files to be stored in a directory named “target” at the root of the ADT distribution.
Data in MS Excel Format
The data and charts used to generate figures for the paper are contained in the following Excel documents. The raw data is available in the PhD. dissertation files distribution below in MS Access format.
- Error over generations.xlsx
- GP Metrics – Decoupled.xlsx
- GP Metrics – Coupled.xlsx
- GP Metrics – GP Only.xlsx
- GP Metrics EMA 1-100.xlsx