contextFold.app
Class Train

java.lang.Object
  extended by contextFold.app.Train

public class Train
extends java.lang.Object

Obtains feature weight parameters of scoring models via a machine-learning procedure.
Usage (from the 'ContextFold' directory):
    "java -cp bin contextFold.app.Train <arg1 name>:X1 <arg2 name>:X2 ...", where Xi is the value to assign for the i-th argument.

Usage examples:
    java -cp bin contextFold.app.Train train:C:/RNAdata/trainData.txt
    java -cp bin contextFold.app.Train train:C:/RNAdata/trainData.txt f:StMedCoHigh
    java -cp bin contextFold.app.Train train:C:/RNAdata/trainData.txt f:trained/Baseline.model iter:6 out:trained/BaselineRetrained improve:0.5

Required argument:
    train - Path for the training data file (must contain structures in addition to sequences).

Main optional arguments (use the 'man' argument for the complete list):
    man - Prints the usage manual.
    f - A scoring model file. Can be either a name of parameterization java file (extending the contextFold.models.FeatureSetter) found in the contextFold.models package, or a path to a previously trained scoring model file (a '.model' file). Default value: 'StHighCoHigh' (using the model defined in the StHighCoHigh.java file, in the contextFold.models package).
    out - Path for the trained model output file (if not specified, a file with the same name as the input file and an the added '.model' extension will be generated in the same directory as the input file).
    iter - Maximum number of iterations of the training procedure (Default: 20).
    shuffle - If true, shuffles the training sequences after each phase of processing the complete set (Allowed values: true (default), false).
    log - Path to a log file. If not specified, log output is printed to screen.
    loss - Loss metric (Allowed values: fMeasure (default), hamming).
    keepIterationWeights - Saving intermediate weights after each iteration (Allowed values: false (default), true).
    improve - improve:X requires that the loss estimation (based on the training set) will improve in at least X percentage (Default: X = 1) after each Y consecutive iterations, where Y is defined by the 'maxNI' argument.
    maxNI - maxNI:Y imposes that if there was no sufficient improvement in the last Y iterations (according to the 'improve' parameter), the process terminates (Default: Y = 3).


Constructor Summary
Train()
           
 
Method Summary
static void main(java.lang.String[] args)
           
 
Methods inherited from class java.lang.Object
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait
 

Constructor Detail

Train

public Train()
Method Detail

main

public static void main(java.lang.String[] args)
                 throws java.io.IOException
Throws:
java.io.IOException