0

我有两个用于特征和权重元素的矩阵。我正在实现一个学习算法。我想更新 arraylist 的元素（表示一个特征样本的向量）。以下是代码。但是我的矩阵元素（向量元素没有）更新了。我也放了样品溶液。更新前后的值不会相同。你能告诉我代码的缺陷在哪里吗？ 

    for(int i =0 ; i< N ; i++){  //N is a large real number
    ArrayList<Double> featureVector = new ArrayList<Double>();
    featureVector = FeatureMatrix.get(i);
    System.out.println("Before::"+ featureVector);
    if(testList.contains(i)){
    for(int j=0 ; j< testList.size(); j++){
        if(i == testList.get(j)){       
        int indexInTestList= j;
        List<Double> subListNextCandidate ;
        subListNextCandidate =  weightVectorNextCandidate.subList((10*indexIntTestList),((10)*(indexInTestList+1))); //clips a portion of member from long list of members
        List<Double> approxWeight = new ArrayList<Double>();
        approxWeight = getApproxWeight(FeatureVector, indexInTestList, FeatureMatrix,WeightMatrix, bias); //approxWeight is a vector of same dimension as of featureVector

        for(int l=0 ; l< 10;l++){                 
            double Update = featureVector.get(l)+ Rate*((subListCandidate.get(l)-approxWeight.get(l))-(lambda*featureVector.get(l)*(1/M)));//M is large real number
            featureVector.set(l,Update);

        }                   
        }
    }
    }

    else{
    for(int l=0 ; l< 10;l++){
        double Update = featureVector.get(l)  -Rate*(lambda*featureVector.get(l)*(1/M));
        featureVector.set(l, Update);
    }                   
    }
    System.out.println("After:::"+ FeatureMatrix.get(i) );  
}

示例输出是::

 Before::[0.04539928251182193, -0.16233604402485394, 0.905018369795912, -1.2817141994528614, 0.7065420460225843, -0.8946090188977665, -1.74892020689701, -2.1539901172158187, 1.8229765478806985, -1.8109945435256574]
 After:::[0.04539928251182193, -0.16233604402485394, 0.905018369795912, -1.2817141994528614, 0.7065420460225843, -0.8946090188977665, -1.74892020689701, -2.1539901172158187, 1.8229765478806985, -1.8109945435256574]
4

2 回答 2

2

我只能想到会发生这种情况的几个合理原因：

  1. 率 == 0
  2. testList.contains(i) 总是假的

我强烈建议使用断点来调试它。至少，在调用 featureVector.set() 的地方放置一个 System.out.println 以确保它被调用过。我猜它从来没有被调用过，因为条件永远不会成为真的。

一定要使用断点，这将是一个救生员......

于 2012-04-24T20:25:10.357 回答
0

的返回类型是testList.get(j)什么？您正在将整数与我怀疑的整数进行比较。这不太可能顺利...

于 2012-05-06T13:16:48.233 回答