The Config folder contains files that allow the user to specify settings that allow ML-Flex to interface with third-party machine-learning software. These settings can be used within any ML-Flex experiment. Below are descriptions of these files and what the configuration values mean.
This file contains information necessary to configure parameters for feature selection/ranking algorithms. Each line in this file contains parameters for a single algorithm. Individual parameters are separated by semicolons. The first value (e.g., "weka_info_gain") indicates a unique key for the algorithm. This same key is referenced from Experiment files. The second value references a "learner" that has been specified in Config/Learner_Templates.txt. The remaining configuration values are parameters that are passed to the learner. These values vary depending on what each learner requires.
This file contains information necessary to configure parameters for classification algorithms. Each line in this file contains parameters for a single classification algorithm. Individual parameters are separated by semicolons. The first value (e.g., "weka_svm_linear") indicates a unique key for the algorithm. This same name is referenced in Experiment files. The second value references a "learner" that has been specified in Config/Learner_Templates.txt. The remaining configuration values are parameters that are passed to the learner. These values vary depending on what each learner requires.
Out of the box, ML-Flex supports mlflex.WekaLearner, which interfaces to the Weka machine-learning package. If configured, ML-Flex should be able interface with any algorithm in Weka. Additionally, ML-Flex contains mlflex.OrangeLearner, which interfaces with the Orange machine-learning framework, mlflex.C5Learner that interfaces with the C5.0 Decision Trees software, and mlflex.RLearner that interfaces with the R statistical package. However, the Orange, C5.0, and R software are not packaged with ML-Flex by default and need to be installed and configured alongside ML-Flex if they are to be used. Paths to these libraries also need to be specified in Classification_Algorithms.txt and/or Feature_Selection_Algorithms.txt. Example files demonstrate how this can be done.
The current implementation of mlflex.RLearner interacts with the Internals/R/Predict.R script. This script currently is designed to interface with the e1071 package. This package must also be installed from within R before running an ML-Flex analysis that invokes this script. Also, please note that Internals/R/Predict.R can be modified to support other R machine-learning packages and/or custom implementations in R.
List of Command-line Arguments
Executing Experiments Across Multiple Computers
Third-party Machine Learning Software