cutils_ext import fix -- allow theano to run on picloud servers (not CVM)

fix bug in gpu inc subtensor optimization and add a unit test

fix bug in gradient when some outputs of a node have different connectio...

git just got rid of this edit)

Misc

(Razvan's branch will introduce a test that catches this bug)

Add erfinv and erfcinv in the doc.

Add test for the "flatten" vector case

This case is similar to what is used in DeepLearningBenchmarks

Fix bug in infer_shape of Flatten

Check the graphviz executable are there.

Even if the pydot Python package is installed, it is not certain that
the graphviz executable are (that is the case with EPD for instance).
In that case, we do not want to use it.

Infershape would incorrectly assign a shape of 1 on the last dimension,
if all dimension were preserved.
In particular, calling .flatten() on some vector would infer a shape of
(1,), which is incorrect.

C code for GpuIncsub, and a new CUDA kernel for one special case

Fix an optimization crash of local_gpu_elemwise_0.

Call to theano.compile.rebuild_collect_shared will clone variable that have owner.