org.nd4j.linalg.factory.Nd4j.getOpFactory()方法的使用及代码示例

x33g5p2x  于2022-01-24 转载在 其他  
字(9.8k)|赞(0)|评价(0)|浏览(146)

本文整理了Java中org.nd4j.linalg.factory.Nd4j.getOpFactory()方法的一些代码示例，展示了Nd4j.getOpFactory()的具体用法。这些代码示例主要来源于Github/Stackoverflow/Maven等平台，是从一些精选项目中提取出来的代码，具有较强的参考意义，能在一定程度帮忙到你。Nd4j.getOpFactory()方法的具体详情如下：
包路径：org.nd4j.linalg.factory.Nd4j
类名称：Nd4j
方法名：getOpFactory

Nd4j.getOpFactory介绍

[英]Get the operation factory
[中]去操作工厂

代码示例

代码示例来源：origin: deeplearning4j/nd4j

  1. public static long getOpNum(String name, Op.Type type) {
  2.   if (type == Op.Type.CUSTOM)
  3.     return Nd4j.getExecutioner().getCustomOperations().get(name.toLowerCase()).getHash();
  4.   else
  5.     return (long) Nd4j.getOpFactory().getOpNumByName(name);
  6. }

代码示例来源：origin: deeplearning4j/nd4j

  1. return (long) Nd4j.getOpFactory().getOpNumByName(name);

代码示例来源：origin: deeplearning4j/nd4j

  1. public INDArray scoreArray(INDArray labels, INDArray preOutput, IActivation activationFn, INDArray mask) {
  2.   if (labels.size(1) != preOutput.size(1)) {
  3.     throw new IllegalArgumentException(
  4.             "Labels array numColumns (size(1) = " + labels.size(1) + ") does not match output layer"
  5.                     + " number of outputs (nOut = " + preOutput.size(1) + ") ");
  6.   }
  7.   INDArray scoreArr;
  8.   //INDArray output = Nd4j.getExecutioner().execAndReturn(Nd4j.getOpFactory().createTransform(activationFn, preOutput.dup()));
  9.   INDArray output = activationFn.getActivation(preOutput.dup(), true);
  10.   scoreArr = output.subi(labels);
  11.   Nd4j.getExecutioner().execAndReturn(Nd4j.getOpFactory().createTransform("abs", scoreArr));
  12.   //Weighted loss function
  13.   if (weights != null) {
  14.     if (weights.length() != output.size(1)) {
  15.       throw new IllegalStateException("Weights vector (length " + weights.length()
  16.               + ") does not match output.size(1)=" + output.size(1));
  17.     }
  18.     scoreArr.muliRowVector(weights);
  19.   }
  20.   if (mask != null) {
  21.     LossUtil.applyMask(scoreArr, mask);
  22.   }
  23.   return scoreArr;
  24. }

代码示例来源：origin: deeplearning4j/nd4j

  1. public INDArray scoreArray(INDArray labels, INDArray preOutput, IActivation activationFn, INDArray mask) {
  2.   if (labels.size(1) != preOutput.size(1)) {
  3.     throw new IllegalArgumentException(
  4.             "Labels array numColumns (size(1) = " + labels.size(1) + ") does not match output layer"
  5.                     + " number of outputs (nOut = " + preOutput.size(1) + ") ");
  6.   }
  7.   INDArray scoreArr;
  8.   //INDArray output = Nd4j.getExecutioner().execAndReturn(Nd4j.getOpFactory().createTransform(activationFn, preOutput.dup()));
  9.   INDArray output = activationFn.getActivation(preOutput.dup(), true);
  10.   scoreArr = output.rsubi(labels).divi(labels);
  11.   Nd4j.getExecutioner().execAndReturn(Nd4j.getOpFactory().createTransform("abs", scoreArr));
  12.   scoreArr.muli(100.0 / labels.size(1));
  13.   //Weighted loss function
  14.   if (weights != null) {
  15.     if (weights.length() != output.size(1)) {
  16.       throw new IllegalStateException("Weights vector (length " + weights.length()
  17.               + ") does not match output.size(1)=" + output.size(1));
  18.     }
  19.     scoreArr.muliRowVector(weights);
  20.   }
  21.   if (mask != null) {
  22.     LossUtil.applyMask(scoreArr, mask);
  23.   }
  24.   return scoreArr;
  25. }

代码示例来源：origin: org.deeplearning4j/nearestneighbor-core

  1. /**
  2.  *
  3.  * @param m1
  4.  * @param m2
  5.  * @return
  6.  */
  7. public double getDistance(Point m1, Point m2) {
  8.   return Nd4j.getExecutioner()
  9.           .execAndReturn(Nd4j.getOpFactory().createAccum(distanceFunction, m1.getArray(), m2.getArray()))
  10.           .getFinalResult().doubleValue();
  11. }

代码示例来源：origin: org.deeplearning4j/nearestneighbor-core

  1. /**
  2.  * Get the distance to the given
  3.  * point from the cluster
  4.  * @param point the point to get the distance for
  5.  * @return
  6.  */
  7. public double getDistanceToCenter(Point point) {
  8.   return Nd4j.getExecutioner().execAndReturn(
  9.           Nd4j.getOpFactory().createAccum(distanceFunction, center.getArray(), point.getArray()))
  10.           .getFinalResult().doubleValue();
  11. }

代码示例来源：origin: org.deeplearning4j/nearestneighbor-core

  1. public void run() {
  2.     try {
  3.       for (int k = clusterIdx + 1, l = clusterSet.getClusterCount(); k < l; k++) {
  4.         Cluster toCluster = clusterSet.getClusters().get(k);
  5.         double distance = Nd4j.getExecutioner()
  6.                 .execAndReturn(Nd4j.getOpFactory().createAccum(
  7.                         clusterSet.getDistanceFunction(),
  8.                         fromCluster.getCenter().getArray(),
  9.                         toCluster.getCenter().getArray()))
  10.                 .getFinalResult().doubleValue();
  11.         info.getDistancesBetweenClustersCenters().put(fromCluster.getId(), toCluster.getId(),
  12.                 distance);
  13.       }
  14.     } catch (Exception e) {
  15.       e.printStackTrace();
  16.     }
  17.   }
  18. });

代码示例来源：origin: org.deeplearning4j/nearestneighbor-core

  1. /**
  2.  *
  3.  * @param cluster
  4.  * @param distanceFunction
  5.  * @return
  6.  */
  7. public static ClusterInfo computeClusterInfos(Cluster cluster, String distanceFunction) {
  8.   ClusterInfo info = new ClusterInfo(cluster.isInverse(), true);
  9.   for (int i = 0, j = cluster.getPoints().size(); i < j; i++) {
  10.     Point point = cluster.getPoints().get(i);
  11.     //shouldn't need to inverse here. other parts of
  12.     //the code should interpret the "distance" or score here
  13.     double distance = Nd4j.getExecutioner()
  14.             .execAndReturn(Nd4j.getOpFactory().createAccum(distanceFunction,
  15.                     cluster.getCenter().getArray(), point.getArray()))
  16.             .getFinalResult().doubleValue();
  17.     info.getPointDistancesFromCenter().put(point.getId(), distance);
  18.     double diff = info.getTotalPointDistanceFromCenter() + distance;
  19.     info.setTotalPointDistanceFromCenter(diff);
  20.   }
  21.   if (!cluster.getPoints().isEmpty())
  22.     info.setAveragePointDistanceFromCenter(info.getTotalPointDistanceFromCenter() / cluster.getPoints().size());
  23.   return info;
  24. }

代码示例来源：origin: org.deeplearning4j/deeplearning4j-nn

  1. public INDArray propUpDerivative(INDArray z) {
  2.   switch (layerConf().getHiddenUnit()) {
  3.     case IDENTITY:
  4.       return Nd4j.getExecutioner()
  5.               .execAndReturn(Nd4j.getOpFactory().createTransform("identity", z).derivative());
  6.     case BINARY:
  7.       return Nd4j.getExecutioner()
  8.               .execAndReturn(Nd4j.getOpFactory().createTransform("sigmoid", z).derivative());
  9.     case GAUSSIAN: {
  10.       Distribution dist = Nd4j.getDistributions().createNormal(z, 1);
  11.       INDArray gaussian = dist.sample(z.shape());
  12.       INDArray derivative = z.mul(-2).mul(gaussian);
  13.       return derivative;
  14.     }
  15.     case RECTIFIED:
  16.       return Nd4j.getExecutioner().execAndReturn(Nd4j.getOpFactory().createTransform("relu", z).derivative());
  17.     case SOFTMAX:
  18.       return Nd4j.getExecutioner()
  19.               .execAndReturn(Nd4j.getOpFactory().createTransform("softmax", z).derivative());
  20.     default:
  21.       throw new IllegalStateException(
  22.               "Hidden unit type should either be binary, gaussian, or rectified linear " + layerId());
  23.   }
  24. }

代码示例来源：origin: org.nd4j/nd4j-api

  1. public INDArray scoreArray(INDArray labels, INDArray preOutput, IActivation activationFn, INDArray mask) {
  2.   if (labels.size(1) != preOutput.size(1)) {
  3.     throw new IllegalArgumentException("Labels array numColumns (size(1) = " + labels.size(1)
  4.             + ") does not match output layer" + " number of outputs (nOut = " + preOutput.size(1)
  5.             + ") ");
  6.     
  7.   }
  8.   INDArray scoreArr;
  9.   //INDArray output = Nd4j.getExecutioner().execAndReturn(Nd4j.getOpFactory().createTransform(activationFn, preOutput.dup()));
  10.   INDArray output = activationFn.getActivation(preOutput.dup(), true);
  11.   scoreArr = output.subi(labels);
  12.   Nd4j.getExecutioner().execAndReturn(Nd4j.getOpFactory().createTransform("abs", scoreArr));
  13.   //Weighted loss function
  14.   if (weights != null) {
  15.     if (weights.length() != output.size(1)) {
  16.       throw new IllegalStateException("Weights vector (length " + weights.length()
  17.               + ") does not match output.size(1)=" + output.size(1));
  18.     }
  19.     scoreArr.muliRowVector(weights);
  20.   }
  21.   if (mask != null) {
  22.     LossUtil.applyMask(scoreArr, mask);
  23.   }
  24.   return scoreArr;
  25. }

代码示例来源：origin: org.deeplearning4j/deeplearning4j-nn

  1. /**
  2.  * Calculates the activation of the visible :
  3.  * sigmoid(v * W + hbias)
  4.  * @param v the visible layer
  5.  * @return the approximated activations of the visible layer
  6.  */
  7. public INDArray propUp(INDArray v, boolean training) {
  8.   INDArray preSig = preOutput(v, training);
  9.   switch (layerConf().getHiddenUnit()) {
  10.     case IDENTITY:
  11.       return preSig;
  12.     case BINARY:
  13.       return sigmoid(preSig);
  14.     case GAUSSIAN:
  15.       Distribution dist = Nd4j.getDistributions().createNormal(preSig, 1);
  16.       preSig = dist.sample(preSig.shape());
  17.       return preSig;
  18.     case RECTIFIED:
  19.       preSig = max(preSig, 0.0);
  20.       return preSig;
  21.     case SOFTMAX:
  22.       return Nd4j.getExecutioner().execAndReturn(Nd4j.getOpFactory().createTransform("softmax", preSig));
  23.     default:
  24.       throw new IllegalStateException(
  25.               "Hidden unit type should either be binary, gaussian, or rectified linear " + layerId());
  26.   }
  27. }

代码示例来源：origin: org.nd4j/nd4j-api

  1. public INDArray scoreArray(INDArray labels, INDArray preOutput, IActivation activationFn, INDArray mask) {
  2.   if (labels.size(1) != preOutput.size(1)) {
  3.     throw new IllegalArgumentException("Labels array numColumns (size(1) = " + labels.size(1)
  4.             + ") does not match output layer" + " number of outputs (nOut = " + preOutput.size(1)
  5.             + ") ");
  6.     
  7.   }
  8.   INDArray scoreArr;
  9.   //INDArray output = Nd4j.getExecutioner().execAndReturn(Nd4j.getOpFactory().createTransform(activationFn, preOutput.dup()));
  10.   INDArray output = activationFn.getActivation(preOutput.dup(), true);
  11.   scoreArr = output.rsubi(labels).divi(labels);
  12.   Nd4j.getExecutioner().execAndReturn(Nd4j.getOpFactory().createTransform("abs", scoreArr));
  13.   scoreArr.muli(100.0 / labels.size(1));
  14.   //Weighted loss function
  15.   if (weights != null) {
  16.     if (weights.length() != output.size(1)) {
  17.       throw new IllegalStateException("Weights vector (length " + weights.length()
  18.               + ") does not match output.size(1)=" + output.size(1));
  19.     }
  20.     scoreArr.muliRowVector(weights);
  21.   }
  22.   if (mask != null) {
  23.     LossUtil.applyMask(scoreArr, mask);
  24.   }
  25.   return scoreArr;
  26. }

代码示例来源：origin: org.deeplearning4j/deeplearning4j-nn

  1. /**
  2.  * Calculates the activation of the hidden:
  3.  * activation(h * W + vbias)
  4.  * @param h the hidden layer
  5.  * @return the approximated output of the hidden layer
  6.  */
  7. public INDArray propDown(INDArray h) {
  8.   INDArray W = getParam(PretrainParamInitializer.WEIGHT_KEY).transpose();
  9.   INDArray vBias = getParam(PretrainParamInitializer.VISIBLE_BIAS_KEY);
  10.   INDArray vMean = h.mmul(W).addiRowVector(vBias);
  11.   switch (layerConf().getVisibleUnit()) {
  12.     case IDENTITY:
  13.       return vMean;
  14.     case BINARY:
  15.       return sigmoid(vMean);
  16.     case GAUSSIAN:
  17.       Distribution dist = Nd4j.getDistributions().createNormal(vMean, 1);
  18.       vMean = dist.sample(vMean.shape());
  19.       return vMean;
  20.     case LINEAR:
  21.       return vMean;
  22.     case SOFTMAX:
  23.       return Nd4j.getExecutioner().execAndReturn(Nd4j.getOpFactory().createTransform("softmax", vMean));
  24.     default:
  25.       throw new IllegalStateException("Visible unit type should either be binary or gaussian " + layerId());
  26.   }
  27. }

代码示例来源：origin: org.deeplearning4j/deeplearning4j-nn

  1. vSample = Nd4j.getExecutioner().execAndReturn(Nd4j.getOpFactory().createTransform("softmax", vProb));
  2. break;

代码示例来源：origin: org.deeplearning4j/deeplearning4j-nn

  1. hSample = Nd4j.getExecutioner().execAndReturn(Nd4j.getOpFactory().createTransform("softmax", hProb));
  2. break;

内容来源于网络，如有侵权，请联系作者删除！

Nd4j

相关文章

    查看更多

    热门标签

    更多
    JavaquerypythonNode开发语言requestUtil数据库Table后端算法LoggerMessageElementParser

    最新文章

    更多

    Nd4j类方法

    • 蜀ICP备13028337号-1 大数据知识库 https://www.saoniuhuo.com © All rights reserved
    • 本站内容来源互联网,如果侵犯您的权益请联系我们删除, 联系方式：448109455@qq.com