为什么在 Keras 指标函数中使用axis=-1? [英] Why use axis=-1 in Keras metrics function?

问题描述

keras 版本:2.0.8

keras version:2.0.8

在一些 Keras 度量函数和损失函数中，使用axis=-1 作为参数.

In some Keras metric functions and loss functions, use axis=-1 as parameter.

例如:

def binary_accuracy(y_true, y_pred):
    return K.mean(K.equal(y_true, K.round(y_pred)), axis=-1)

就我而言:

y_true 的形状:(4,256,256,2)

shape of y_true:(4,256,256,2)

y_pred 的形状:(4,256,256,2)

shape of y_pred:(4,256,256,2)

因此，binary_accuracy(y_true, y_pred) 应该返回一个 shape=(4,256,256) 的张量，而不是一个标量张量.

So, binary_accuracy(y_true, y_pred) should return a tensor with shape=(4,256,256) instead of a scalar tensor.

但是当使用 binary_accuracy 作为度量函数时:

But when use binary_accuracy as metric function:

model.compile(optimizer=adam, loss=keras.losses.binary_crossentropy, metrics=[binary_accuracy])

日志仍然将 binary_accuracy 打印为标量，这让我很困惑.

The log still prints binary_accuracy as scalar,which confused me a lot.

keras 是否对 binary_accuracy 函数的返回做了一些特殊处理?

Does keras do some special on the return of binary_accuracy function?

纪元 11/300

0s - 损失:0.4158 - binary_accuracy:0.9308 - val_loss:0.4671 -val_binary_accuracy: 0.7767

0s - loss: 0.4158 - binary_accuracy: 0.9308 - val_loss: 0.4671 - val_binary_accuracy: 0.7767

推荐答案

这里是您要找的内容，training_utils.py:

Here's what you're looking for, inside training_utils.py:

def weighted(y_true, y_pred, weights, mask=None):
    """Wrapper function.
    # Arguments
        y_true: `y_true` argument of `fn`.
        y_pred: `y_pred` argument of `fn`.
        weights: Weights tensor.
        mask: Mask tensor.
    # Returns
        Scalar tensor.
    """
    # score_array has ndim >= 2
    score_array = fn(y_true, y_pred)
    if mask is not None:
        # Cast the mask to floatX to avoid float64 upcasting in Theano
        mask = K.cast(mask, K.floatx())
        # mask should have the same shape as score_array
        score_array *= mask
        #  the loss per batch should be proportional
        #  to the number of unmasked samples.
        score_array /= K.mean(mask) + K.epsilon()

    # apply sample weighting
    if weights is not None:
        # reduce score_array to same ndim as weight array
        ndim = K.ndim(score_array)
        weight_ndim = K.ndim(weights)
        score_array = K.mean(score_array,
                             axis=list(range(weight_ndim, ndim)))
        score_array *= weights
        score_array /= K.mean(K.cast(K.not_equal(weights, 0), K.floatx()))
    return K.mean(score_array)
return weighted

度量函数由 score_array = fn(y_true, y_pred) 调用(它是一个嵌套函数，fn 在外部函数中定义).该数组在最后一行 return K.mean(score_array) 中取平均值.这就是为什么您看到的是标量指标而不是张量.中间的线只是在必要时引入掩码和权重.

The metric function is called by score_array = fn(y_true, y_pred) (it's a nested function and fn is defined in the outer function). This array is averaged in the last line return K.mean(score_array). That's why you're seeing scalar metrics instead of tensors. The lines in between are just to introduce masks and weights if necessary.

这篇关于为什么在 Keras 指标函数中使用axis=-1?的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！