K.local_conv1d

• 关键输入
  1. inputs: 3D tensor with shape: (batch_size, steps, input_dim)
  2. kernel: the unshared weight for convolution,
     with shape (output_length, feature_dim, filters)
  3. kernel_size: a tuple of a single integer,
     specifying the length of the 1D convolution window
  4. strides: a tuple of a single integer,
     specifying the stride length of the convolution
• 注意点
  1. steps表示kernel要在该维度上移动
  2. steps维度的大小应该等于 (output_length-1)*stride + kernel_size
  3. feature_dim 应该是被 batch_size * kernel_size * input_dim 整除

K.local_conv2d

• 输入
  1. inputs: 4D tensor with shape:
     (batch_size, filters, new_rows, new_cols)
     if data_format='channels_first'
     or 4D tensor with shape:
     (batch_size, new_rows, new_cols, filters)
     if data_format='channels_last'.
  2. kernel: the unshared weight for convolution,
     with shape (output_items, feature_dim, filters)
  3. kernel_size: a tuple of 2 integers, specifying the
     width and height of the 2D convolution window.
  4. strides: a tuple of 2 integers, specifying the strides
     of the convolution along the width and height.
  5. output_shape: a tuple with (output_row, output_col)
  6. data_format: the data format, channels_first or channels_last
• output_row, output_col的要求和local_conv1d中output_length的要求类似
• kernel 中的 feature_dim 应该可以被 batch_size * filters * kernel_size[0] * kernel_size[1] 整除
• kernel中的output_items = output_shape[0] * output_shape[1]
• kernel中的feature_dim的含义是什么？难道要等于 kernel_size[0]*kernel_size[0]
• inputs中的 filters 指输入filter
• kernel中的filters 指输出filter, 输入中的filter和输出中的filter 不必相等
• 要想输出的batch_size 等于 输入的batch_size, kernel中的feature_dim应该等等于kernel_size[0] * kernel_size[1] * 输入中的filters
• 返回
  A 4d tensor with shape:
  (batch_size, filters, new_rows, new_cols)
  if data_format='channels_first'
  or 4D tensor with shape:
  `(batch_size, new_rows, new_cols, filters)
  if data_format='channels_last'.