In keras 2.0 tutorial I don’t see an mode option in BatchNormalization(). The stuff missing is below,
mode: integer, 0, 1 or 2.
0: feature-wise normalization. Each feature map in the input will be normalized separately. The axis on which to normalize is specified by the axis argument. Note that if the input is a 4D image tensor using Theano conventions (samples, channels, rows, cols) then you should set axis to 1 to normalize along the channels axis. During training we use per-batch statistics to normalize the data, and during testing we use running averages computed during the training phase.
1: sample-wise normalization. This mode assumes a 2D input.
2: feature-wise normalization, like mode 0, but using per-batch statistics to normalize the data during both testing and training.
As the above was there in keras 1.2 but seems to be omitted in keras 2. Also setting axis = 1 in Convolution layers seems to produce better results in Tensorflow backed Environment than which is recommended.
Any ideas anybody want to share ?