/******************************************************************************* * Copyright 2015 Analog Devices, Inc. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. ********************************************************************************/ package com.analog.lyric.dimple.solvers.sumproduct.sampledfactor; import com.analog.lyric.dimple.model.variables.Real; import com.analog.lyric.dimple.solvers.core.parameterizedMessages.NormalParameters; import com.analog.lyric.dimple.solvers.gibbs.GibbsOptions; import com.analog.lyric.dimple.solvers.gibbs.GibbsReal; import com.analog.lyric.dimple.solvers.sumproduct.SumProductNormalEdge; import com.analog.lyric.util.misc.Internal; import cern.colt.list.DoubleArrayList; /** * * @since 0.08 * @author Christopher Barber * @category internal */ @Internal class SumProductSampledNormalEdge extends SumProductNormalEdge implements ISumProductSampledEdge<NormalParameters> { private final GibbsReal _svar; SumProductSampledNormalEdge(GibbsReal svar) { super(); _svar = svar; setVariableInputUniform(); } @Override public void setFactorToVarDirection() { _svar.setOption(GibbsOptions.saveAllSamples, true); setVariableInputUniform(); } @Override public void setVarToFactorDirection() { _svar.setOption(GibbsOptions.saveAllSamples, false); Real var = _svar.getModelObject(); if (!var.hasFixedValue()) // Only set the input if there isn't already a fixed value { NormalParameters inputMessage = varToFactorMsg; if (varToFactorMsg.getPrecision() == 0) { var.setPrior(null); // If zero precision, then set the input to null to avoid numerical issues } else { var.setPrior(inputMessage); } } } @Override public void setFactorToVarMsgFromSamples() { final NormalParameters outputMessage = factorToVarMsg; // Get the raw sample array to avoid making a copy; this is unsafe, so be careful not to modify it @SuppressWarnings("null") DoubleArrayList sampleValues = _svar._getSampleArrayUnsafe(); @SuppressWarnings("null") int numSamples = sampleValues.size(); // For all sample values, compute the output message double sum = 0; double sumsq = 0; for (int i = 0; i < numSamples; i++) { double tmp = sampleValues.get(i); if (Double.isInfinite(tmp) || Double.isNaN(tmp)) { outputMessage.setNull(); return; } sum += tmp; sumsq += tmp*tmp; } double mean = sum / numSamples; double variance = (sumsq - sum*mean) / (numSamples - 1); outputMessage.setMean(mean); outputMessage.setVariance(variance); } private final void setVariableInputUniform() { Real var = _svar.getModelObject(); if (!var.hasFixedValue()) // Only set the input if there isn't already a fixed value { var.setPrior(null); } } }