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

x33g5p2x  于2022-01-24 转载在 其他  
字(6.6k)|赞(0)|评价(0)|浏览(191)

本文整理了Java中org.nd4j.linalg.factory.Nd4j.ones()方法的一些代码示例，展示了Nd4j.ones()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.ones()方法的具体详情如下：
包路径：org.nd4j.linalg.factory.Nd4j
类名称：Nd4j
方法名：ones

Nd4j.ones介绍

[英]Creates a row vector with the specified number of columns
[中]创建具有指定列数的行向量

代码示例

代码示例来源：origin: deeplearning4j/nd4j

  1. /**
  2.  * Ones like
  3.  *
  4.  * @param arr the array to create the ones like
  5.  * @return ones in the shape of the given array
  6.  */
  7. public static INDArray onesLike(INDArray arr) {
  8.   return ones(arr.shape());
  9. }

代码示例来源：origin: deeplearning4j/dl4j-examples

  1. INDArray ones = Nd4j.ones(nRows, nColumns);

代码示例来源：origin: deeplearning4j/nd4j

  1. public double getGradient(double gradient, int column, int[] shape) {
  2.   boolean historicalInitialized = false;
  3.   if (this.historicalGradient == null) {
  4.     this.historicalGradient = Nd4j.ones(shape);
  5.     historicalInitialized = true;
  6.   }
  7.   double sqrtHistory = !historicalInitialized ? Math.sqrt(historicalGradient.getDouble(column))
  8.           : historicalGradient.getDouble(column);
  9.   double learningRates = learningRate / (sqrtHistory + epsilon);
  10.   double adjustedGradient = gradient * (learningRates);
  11.   historicalGradient.putScalar(column, historicalGradient.getDouble(column) + gradient * gradient);
  12.   numIterations++;
  13.   //ensure no zeros
  14.   return adjustedGradient;
  15. }

代码示例来源：origin: deeplearning4j/dl4j-examples

  1. System.out.println(allZeros);
  2. INDArray allOnes = Nd4j.ones(nRows, nColumns);
  3. System.out.println("\nNd4j.ones(nRows, nColumns)");
  4. System.out.println(allOnes);
  5. INDArray threeDimArray = Nd4j.ones(3,4,5);      //3x4x5 INDArray
  6. INDArray fourDimArray = Nd4j.ones(3,4,5,6);     //3x4x5x6 INDArray
  7. INDArray fiveDimArray = Nd4j.ones(3,4,5,6,7);   //3x4x5x6x7 INDArray
  8. System.out.println("\n\n\nCreating INDArrays with more dimensions:");
  9. System.out.println("3d array shape:         " + Arrays.toString(threeDimArray.shape()));

代码示例来源：origin: deeplearning4j/nd4j

  1. @Override
  2. public INDArray computeGradient(INDArray labels, INDArray preOutput, IActivation activationFn, INDArray mask) {
  3.   if (labels.size(1) != preOutput.size(1)) {
  4.     throw new IllegalArgumentException(
  5.         "Labels array numColumns (size(1) = " + labels.size(1) + ") does not match output layer"
  6.             + " number of outputs (nOut = " + preOutput.size(1) + ") ");
  7.   }
  8.   final INDArray grad = Nd4j.ones(labels.shape());
  9.   calculate(labels, preOutput, activationFn, mask, null, grad);
  10.   return grad;
  11. }

代码示例来源：origin: deeplearning4j/nd4j

  1. @Override
  2. public Pair<INDArray, INDArray> backprop(INDArray in, INDArray epsilon) {
  3.   INDArray dLdz = Nd4j.ones(in.shape());
  4.   BooleanIndexing.replaceWhere(dLdz, alpha, Conditions.lessThanOrEqual(0.0));
  5.   dLdz.muli(epsilon);
  6.   return new Pair<>(dLdz, null);
  7. }

代码示例来源：origin: deeplearning4j/nd4j

  1. @Override
  2. public Pair<Double, INDArray> computeGradientAndScore(INDArray labels,
  3.                            INDArray preOutput, IActivation activationFn, INDArray mask, boolean average) {
  4.   final INDArray scoreArr = Nd4j.create(labels.size(0), 1);
  5.   final INDArray grad = Nd4j.ones(labels.shape());
  6.   calculate(labels, preOutput, activationFn, mask, scoreArr, grad);
  7.   double score = scoreArr.sumNumber().doubleValue();
  8.   if (average)
  9.     score /= scoreArr.size(0);
  10.   return new Pair<>(score, grad);
  11. }

代码示例来源：origin: deeplearning4j/nd4j

  1. public static INDArray mergePerOutputMasks2d(long[] outShape, INDArray[] arrays, INDArray[] masks) {
  2.   val numExamplesPerArr = new long[arrays.length];
  3.   for (int i = 0; i < numExamplesPerArr.length; i++) {
  4.     numExamplesPerArr[i] = arrays[i].size(0);
  5.   }
  6.   INDArray outMask = Nd4j.ones(outShape); //Initialize to 'all present' (1s)
  7.   int rowsSoFar = 0;
  8.   for (int i = 0; i < masks.length; i++) {
  9.     long thisRows = numExamplesPerArr[i]; //Mask itself may be null -> all present, but may include multiple examples
  10.     if (masks[i] == null) {
  11.       continue;
  12.     }
  13.     outMask.put(new INDArrayIndex[] {NDArrayIndex.interval(rowsSoFar, rowsSoFar + thisRows),
  14.             NDArrayIndex.all()}, masks[i]);
  15.     rowsSoFar += thisRows;
  16.   }
  17.   return outMask;
  18. }

代码示例来源：origin: deeplearning4j/dl4j-examples

  1. INDArray values = Nd4j.ones(3,4);
  2. SDVariable variable = sd.var("myVariable", values);

代码示例来源：origin: deeplearning4j/nd4j

  1. public AdaGrad createSubset(int index) {
  2.     if (historicalGradient == null)
  3.       this.historicalGradient = Nd4j.ones(shape);
  5.     if (Shape.isMatrix(shape)) {
  6.       AdaGrad a = new AdaGrad(1, historicalGradient.columns());
  7.       //grab only the needed elements
  8.       INDArray slice = historicalGradient.slice(index).dup();
  9.       a.historicalGradient = slice;
  10.       a.setLearningRate(learningRate);
  11.       return a;
  12.     } else {
  13.       AdaGrad a = new AdaGrad(1, 1);
  14.       //grab only the needed elements
  15.       INDArray slice = Nd4j.scalar(historicalGradient.getDouble(index));
  16.       a.historicalGradient = slice;
  17.       a.setLearningRate(learningRate);
  18.       return a;
  19.     }
  20.   }
  21. }

代码示例来源：origin: deeplearning4j/nd4j

  1. /**
  2.  * Merge the vectors and append a bias.
  3.  * Each vector must be either row or column vectors.
  4.  * An exception is thrown for inconsistency (mixed row and column vectors)
  5.  *
  6.  * @param vectors the vectors to merge
  7.  * @return the merged ndarray appended with the bias
  8.  */
  9. @Override
  10. public INDArray appendBias(INDArray... vectors) {
  11.   int size = 0;
  12.   for (INDArray vector : vectors) {
  13.     size += vector.rows();
  14.   }
  15.   INDArray result = Nd4j.create(size + 1, vectors[0].columns());
  16.   int index = 0;
  17.   for (INDArray vector : vectors) {
  18.     INDArray put = toFlattened(vector, Nd4j.ones(1));
  19.     result.put(new INDArrayIndex[] {NDArrayIndex.interval(index, index + vector.rows() + 1),
  20.             NDArrayIndex.interval(0, vectors[0].columns())}, put);
  21.     index += vector.rows();
  22.   }
  23.   return result;
  24. }

代码示例来源：origin: deeplearning4j/dl4j-examples

  1. INDArray values = Nd4j.ones(3,4);
  2. var3.setArray(values);

代码示例来源：origin: deeplearning4j/nd4j

  1. public INDArray adjustMasks(INDArray label, INDArray labelMask, int minorityLabel, double targetDist) {
  2.     labelMask = Nd4j.ones(label.size(0), label.size(2));

代码示例来源：origin: deeplearning4j/nd4j

  1. final Double locNormFactor = normFactor.getDouble(i);
  2. final INDArray operandA = Nd4j.ones(shape[1], shape[0]).mmul(locCfn);
  3. final INDArray operandB = operandA.transpose();

代码示例来源：origin: deeplearning4j/nd4j

  1. INDArray mask = (needMask && maskRank != 3 ? Nd4j.ones(totalExamples, maxLength) : null);

代码示例来源：origin: deeplearning4j/dl4j-examples

  1. print("One dimensional zeros", oneDZeros);
  2. INDArray threeByFourOnes = Nd4j.ones(3, 4);
  3. print("3x4 ones", threeByFourOnes);

代码示例来源：origin: org.nd4j/nd4j-api

  1. /**
  2.  * Ones like
  3.  *
  4.  * @param arr the array to create the ones like
  5.  * @return ones in the shape of the given array
  6.  */
  7. public static INDArray onesLike(INDArray arr) {
  8.   return ones(arr.shape());
  9. }

代码示例来源：origin: improbable-research/keanu

  1. public static INDArray ones(long[] shape, DataBuffer.Type bufferType) {
  2.   Nd4j.setDataType(bufferType);
  3.   switch (shape.length) {
  4.     case 0:
  5.       return scalar(1.0, bufferType);
  6.     case 1:
  7.       return reshapeToVector(Nd4j.ones(shape));
  8.     default:
  9.       return Nd4j.ones(shape);
  10.   }
  11. }

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

  1. private INDArray rowOfLogTransitionMatrix(int k) {
  2.   INDArray row = Nd4j.ones(1, states).muli(logOfDiangnalTProb);
  3.   row.putScalar(k, logMetaInstability);
  4.   return row;
  5. }

代码示例来源：origin: org.nd4j/nd4j-api

  1. @Override
  2. public Pair<INDArray, INDArray> backprop(INDArray in, INDArray epsilon) {
  3.   INDArray dLdz = Nd4j.ones(in.shape());
  4.   BooleanIndexing.replaceWhere(dLdz, alpha, Conditions.lessThanOrEqual(0.0));
  5.   dLdz.muli(epsilon);
  6.   return new Pair<>(dLdz, null);
  7. }

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