Examples of com.heatonresearch.aifh.error.ErrorCalculation

com.heatonresearch.aifh.error.ErrorCalculation
An error calculation metric calculates the difference between two vector sets. One vector set will be the ideal expected output from a Machine Learning Algorithm. The other vector set is the actual output. Training a Machine Learning Algorithm typically involves minimizing this error.
Error calculation metrics are very similar to distance metrics. However, an error calculation metric operates over a set of vectors, whereas a distance metric operates over just two vectors.

    /**
     * {@inheritDoc}
     */
    @Override
    public double calculateScore(final MLMethod algo) {
        ErrorCalculation ec = this.errorCalc.create();


        final RegressionAlgorithm ralgo = (RegressionAlgorithm) algo;
        final Genome genome = (Genome)ralgo;


        if( genome.size()>this.maxLength ) {
            return Double.POSITIVE_INFINITY;
        }


        // evaulate
        ec.clear();
        for (final BasicData pair : this.trainingData) {
            final double[] output = ralgo.computeRegression(pair.getInput());
            ec.updateError(output, pair.getIdeal(), 1.0);
        }


        return ec.calculate();
    }

View Full Code Here


    @Test
    public void testGeneral() {
        final List<BasicData> training = BasicData.convertArrays(TEST_INPUT, TEST_IDEAL);
        final ScoreRegressionData score = new ScoreRegressionData(training);
        final ErrorCalculation ec = new ErrorCalculationSSE();
        score.setErrorCalc(ec);
        assertEquals(ec, score.getErrorCalc());
    }

View Full Code Here

     */
    public void process() {


        final NumberFormat nf = NumberFormat.getInstance();


        final ErrorCalculation calcESS = new ErrorCalculationSSE();
        final ErrorCalculation calcMSE = new ErrorCalculationMSE();
        final ErrorCalculation calcRMS = new ErrorCalculationRMS();


        final DataHolder smallErrors = generate(SEED, ROWS, COLS, LOW, HIGH, 0.1);
        final DataHolder mediumErrors = generate(SEED, ROWS, COLS, LOW, HIGH, 0.5);
        final DataHolder largeErrors = generate(SEED, ROWS, COLS, LOW, HIGH, 1.0);
        final DataHolder hugeErrors = generate(SEED, ROWS, COLS, LOW, HIGH, 10.0);

View Full Code Here

    /**
     * {@inheritDoc}
     */
    @Override
    public double calculateScore(final MLMethod algo) {
        ErrorCalculation ec = this.errorCalc.create();


        final RegressionAlgorithm ralgo = (RegressionAlgorithm) algo;
        // evaulate
        ec.clear();
        for (final BasicData pair : this.trainingData) {
            final double[] output = ralgo.computeRegression(pair.getInput());
            ec.updateError(output, pair.getIdeal(), 1.0);
        }


        return ec.calculate();
    }

View Full Code Here

TOP

Related Classes of com.heatonresearch.aifh.error.ErrorCalculation

com.heatonresearch.aifh.examples.error.EvaluateErrors

com.heatonresearch.aifh.examples.gp.ScoreSmallExpression

com.heatonresearch.aifh.learning.score.ScoreRegressionData

com.heatonresearch.aifh.learning.score.TestScoreRegressionData

All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.