public double[] pValues() {
double[] res = zValues();
RealDistribution rd =
_dispersionEstimated
? new TDistribution(_training_metrics.residualDegreesOfFreedom())
: new NormalDistribution();
for (int i = 0; i < res.length; ++i) res[i] = 2 * rd.cumulativeProbability(-Math.abs(res[i]));
return res;
}
public static void main(String[] args) {
// TODO Auto-generated method stub
TextParser parser = new TextParser();
IntegerDistribution id =
(IntegerDistribution) parser.parseText("uniform()", TextParser.INTEGER);
for (int i = 1; i < 101; i++) {
System.out.print(", " + id.sample());
}
System.out.print("\n");
RealDistribution rd = (RealDistribution) parser.parseText("exponential(4,0)", TextParser.REAL);
for (int i = 1; i < 101; i++) {
System.out.print(", " + rd.sample());
}
}
/**
* Initialize weights. This includes steps for doing a random initialization of W as well as the
* vbias and hbias
*/
protected void initWeights() {
if (this.nVisible < 1)
throw new IllegalStateException("Number of visible can not be less than 1");
if (this.nHidden < 1)
throw new IllegalStateException("Number of hidden can not be less than 1");
if (this.dist == null)
dist =
new NormalDistribution(rng, 0, .01, NormalDistribution.DEFAULT_INVERSE_ABSOLUTE_ACCURACY);
/*
* Initialize based on the number of visible units..
* The lower bound is called the fan in
* The outer bound is called the fan out.
*
* Below's advice works for Denoising AutoEncoders and other
* neural networks you will use due to the same baseline guiding principles for
* both RBMs and Denoising Autoencoders.
*
* Hinton's Guide to practical RBMs:
* The weights are typically initialized to small random values chosen from a zero-mean Gaussian with
* a standard deviation of about 0.01. Using larger random values can speed the initial learning, but
* it may lead to a slightly worse final model. Care should be taken to ensure that the initial weight
* values do not allow typical visible vectors to drive the hidden unit probabilities very close to 1 or 0
* as this significantly slows the learning.
*/
if (this.W == null) {
this.W = DoubleMatrix.zeros(nVisible, nHidden);
for (int i = 0; i < this.W.rows; i++)
this.W.putRow(i, new DoubleMatrix(dist.sample(this.W.columns)));
}
this.wAdaGrad = new AdaGrad(this.W.rows, this.W.columns);
if (this.hBias == null) {
this.hBias = DoubleMatrix.zeros(nHidden);
/*
* Encourage sparsity.
* See Hinton's Practical guide to RBMs
*/
// this.hBias.subi(4);
}
this.hBiasAdaGrad = new AdaGrad(hBias.rows, hBias.columns);
if (this.vBias == null) {
if (this.input != null) {
this.vBias = DoubleMatrix.zeros(nVisible);
} else this.vBias = DoubleMatrix.zeros(nVisible);
}
this.vBiasAdaGrad = new AdaGrad(vBias.rows, vBias.columns);
}
private int getOccupationTime() {
double sample = realDistribution.sample();
if (sample > longestParkingTime) {
sample = longestParkingTime;
}
if (sample < 0) {
sample = 0;
}
return (int) sample + parkingSlotProvider.getObject(parkingSlot).getDistance() * 2;
}
public INDArray genGaussianNoise(int id) {
if (!currentNoise.containsKey(id)) {
INDArray zeroMean = Nd4j.zeros(inputSize, outputSize);
currentNoise.put(id, Sampling.normal(RandomUtils.getRandomGenerator(id), zeroMean, noiseVar));
} else {
RealDistribution reals =
new NormalDistribution(
RandomUtils.getRandomGenerator(id),
0,
noiseVarSqrt,
NormalDistribution.DEFAULT_INVERSE_ABSOLUTE_ACCURACY);
INDArrayUtils.shiftLeft(
currentNoise.get(id),
inputSize,
outputSize,
RandomUtils.getRandomGenerator(id).nextInt(inputSize * outputSize),
reals.sample());
}
// currentNoise = Sampling.normal(RandomUtils.getRandomGenerator(id), zeroMean, noiseVar);
return currentNoise.get(id);
}
public double nextDistributedDouble() {
return normal.inverseCumulativeProbability(random.nextDouble());
}