Question

我目前正在尝试重新创建在this paper here中作为神经网络项目的一部分进行的实验。

从this post中得到一些启发，我编写了此python函数，以便能够使用与本文中指定的快速梯度符号方法相同的广告训练过程。此函数通过Keras库利用Python togeather的“高阶函数”功能：

def adveserial_loss(loss_function, model, x ,eps = 0.01):
    def loss_wrapper(y_true, y_pred):
        loss = loss_function(y_true, y_pred)t
        grad = K.gradients(loss, x)
        xb = x + eps * K.sign(grad)
        advloss = loss_function(y_true, model(xb))
        return K.mean(loss+advloss,axis=-1)
    return loss_wrapper

然后我将其用作我的损失（或能量）函数（例如，使用绝对交叉熵）：

model.compile(loss=adveserial_loss(loss_function = categorical_crossentropy, model, model.inputs)

但是，运行这样的代码，会出现以下错误：

Traceback (most recent call last):
  File "src/etap3sim.py", line 48, in <module>
    model.compile(optimizer="adam", loss=ann.adveserial_loss(klosses.categorical_crossentropy,model,model.inputs), metrics=["accuracy"])
  File "/home/xiaoming/.local/lib/python3.7/site-packages/keras/engine/training.py", line 342, in compile
    sample_weight, mask)
  File "/home/xiaoming/.local/lib/python3.7/site-packages/keras/engine/training_utils.py", line 404, in weighted
    score_array = fn(y_true, y_pred)
  File "/home/xiaoming/gitRepos/TIFX05-MScThesis-HenryYang/src/ANN/loss_functions.py", line 29, in loss_wrapper
    advloss = loss_function(y_true, model(xb))
  File "/home/xiaoming/.local/lib/python3.7/site-packages/keras/engine/base_layer.py", line 457, in __call__
    output = self.call(inputs, **kwargs)
  File "/home/xiaoming/.local/lib/python3.7/site-packages/keras/engine/network.py", line 564, in call
    output_tensors, _, _ = self.run_internal_graph(inputs, masks)
  File "/home/xiaoming/.local/lib/python3.7/site-packages/keras/engine/network.py", line 721, in run_internal_graph
    layer.call(computed_tensor, **kwargs))
  File "/home/xiaoming/.local/lib/python3.7/site-packages/keras/layers/convolutional.py", line 171, in call
    dilation_rate=self.dilation_rate)
  File "/home/xiaoming/.local/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py", line 3650, in conv2d
    data_format=tf_data_format)
  File "/home/xiaoming/.local/lib/python3.7/site-packages/tensorflow/python/ops/nn_ops.py", line 893, in convolution
    name=name)
  File "/home/xiaoming/.local/lib/python3.7/site-packages/tensorflow/python/ops/nn_ops.py", line 955, in convolution_internal
    strides = _get_sequence(strides, n, channel_index, "strides")
  File "/home/xiaoming/.local/lib/python3.7/site-packages/tensorflow/python/ops/nn_ops.py", line 72, in _get_sequence
    name, n, n + 2, current_n))
ValueError: strides should be of length 1, 3 or 5 but was 2

有人遇到相同的错误或类似的东西吗？关于如何解决此问题的任何好的建议？任何帮助将不胜感激！

Keras和ConvNets的快速梯度符号方法

0 个答案: