* add cudnn_conv_op

* WIP

* update

* update

* fix grad check

* use platform::memory

* add support group for cudnn

* update

* follow comments

* fix onlycpu build

* update cuda define

* follow comments

* follow comments

* merge with updates

* fix compile error

* follow comments

* follow comments

* Adding thresholded_relu op
* Adding test for thresholded relu op

* init

* unify CopyFrom interface

* fix gpu build error

* fix bug in tensor_py.h

* refine code comments and add TODO list

* fix conflicts in FeedOp and FetchOp

* implementing softplus

* small fix

* small fix

* small fix

* small fix

* Implemented the hardShrink activation

* Fixing the unit test

* Add numerically-stable logsigmoid activation

* Add softshrink operator

* Adjust relative tolerance for grad-check

* Address review comments

* conv_shift_op: initial implementation using Eigen

Limitations:
- both gradient outputs must be specified and are always computed
- explicit for loops => could be optimized in various ways
  (e.g., different memory layout)

* conv shift - gradient fixes

fix case when not all output gradients desired

* conv shift: minor cleanup

* conv shift - more minor cleanup

* conv shift: clean up & initial GPU implementation

* fix rebase issue

Separate the declarations and implementation of the PoolOp and PoolMaker class in order to reuse in pool_cudnn

It will significantly reduce binary size. It is useful for mobile
deployment.

* Implementing the Adamax optimizer step operator
* Adding unit tests for adamax_op

* Changing learning rate and time step to inputs from attributes

* Changing learning rate and time step to input(tensors)

* Making the Adamax operator conform to naming convention

* Removing Tensor<float> from comments

* Rectifying the Adamax implementation

* Changing Unit Test values and adding comments

* Changing Unit Test to test multiple steps