Updated as of 06/21/2018

1. conda install numpy scipy mkl-service libpython m2w64-toolchain #dependencies 

   (Only the last 'libpython' and 'm2w64-toolchain' are needed actually for the anaconda distribtuion）(see https://github.com/Theano/Theano/issues/6614)

2. pip install --upgrade --no-deps git+git://github.com/Theano/Theano.git

3. conda install pygpu 

  #conda install -c conda-forge pygpu #install pygpu-0.7.6

(Note: theano does not support python 3.6)

1. import theano #test theano

   
   Using cuDNN version 5110 on context None
   Mapped name None to device cuda: GeForce 940M (0000:01:00.0)

Note .theanorc:

[global]
floatX = float32
device = cuda

[nvcc]
fastmath = True
compiler_bindir = C:\Program Files (x86)\Microsoft Visual Studio 14.0\VC\bin

[dnn]
library_path = C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\lib\x64
include_path = C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\include
enabled = True

As MILA will stop develop theano (https://groups.google.com/forum/#!msg/theano-users/7Poq8BZutbY/rNCIfvAEAwAJ), consider switching to Tensorflow as back engine next.