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

Nd4j.averageAndPropagate介绍

[英]This method averages input arrays, and returns averaged array. On top of that, averaged array is propagated to all input arrays
[中]此方法平均输入数组，并返回平均数组。除此之外，平均数组被传播到所有输入数组

代码示例

代码示例来源：origin: org.deeplearning4j/deeplearning4j-parallel-wrapper_2.11

private double getScore(AtomicInteger locker) {
  wasAveraged = true;
  double score = 0.0;
  List<INDArray> params = new ArrayList<>();
  for (int cnt = 0; cnt < workers && cnt < locker.get(); cnt++) {
    params.add(zoo[cnt].getModel().params());
    score += zoo[cnt].getModel().score();
  }
  Nd4j.averageAndPropagate(null, params);
  score /= Math.min(workers, locker.get());
  // TODO: improve this
  if (reportScore)
    log.info("Averaged score: " + score);
  return score;
}

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

private double getScore(AtomicInteger locker) {
  wasAveraged = true;
  double score = 0.0;
  List<INDArray> params = new ArrayList<>();
  for (int cnt = 0; cnt < workers && cnt < locker.get(); cnt++) {
    params.add(zoo[cnt].getModel().params());
    score += zoo[cnt].getModel().score();
  }
  Nd4j.averageAndPropagate(null, params);
  score /= Math.min(workers, locker.get());
  // TODO: improve this
  if (reportScore)
    log.info("Averaged score: " + score);
  return score;
}

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

Nd4j.averageAndPropagate(null, updaters);
  batchSize += workerModel.batchSize();
Nd4j.averageAndPropagate(null, updaters);

代码示例来源：origin: org.deeplearning4j/deeplearning4j-parallel-wrapper_2.11

Nd4j.averageAndPropagate(null, updaters);
  batchSize += workerModel.batchSize();
Nd4j.averageAndPropagate(null, updaters);