Allow a None batch dimension in SPSA for compatibility with generate_…

[juesato]

…np()

Also add a test that the generate_np() interface works properly

Addresses #502

[catherio]

cc @nottombrown

[nottombrown]

LGTM after the small changes that I mentioned.

I'm a bit confused by why we have the "batch size must be 1" constraint here in the first place. If it's for performance reasons, perhaps we could put each sample in a tf.while loop?

Doesn't seem super important though, and these changes are definitely improvements so they should be merged in.

[nottombrown]

I think we're [N, H, W, C] now?

[juesato]

The TF implementation still only supports batch size 1. It's implemented this way just for simplicity - when I used it I didn't need more than batch size 1. There's wouldn't really be any gain in efficiency with batching, since SPSA is internally batching the function evaluations of f(x + delta) for all the different delta values at each iteration.

The TF implementation could of course support a tf.while_loop on the outside to loop over the batch dimension, but that seems overly complex to me, since the computation is really handling each image separately, and it's easy to handle outside TF. I think it makes sense for the numpy interface to support batches, so I changed that.

[nottombrown]

I agree with that logic. lgtm!

[nottombrown]

Can we allow arbitrary batch sizes and then iterate through them in the function here? It seems like the "x_val must be a batch of a single image" constraint will probably be surprising to people.

[juesato]

Done.

[juesato]

Added support for batch calls to generate_np().

Pure TF interface doesn't support batched calls (see comment above).

[resya173]

Hi

[npapernot]

The pep8 checker is failing:

======================================================================
FAIL: Test if pep8 is respected.
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/travis/miniconda/envs/test-environment/lib/python2.7/site-packages/nose/case.py", line 197, in runTest
    self.test(*self.arg)
  File "/home/travis/build/tensorflow/cleverhans/cleverhans/devtools/tests/test_format.py", line 45, in test_format_pep8
    raise AssertionError("PEP8 Format not respected")
AssertionError: PEP8 Format not respected
-------------------- >> begin captured stdout << ---------------------
/home/travis/build/tensorflow/cleverhans/cleverhans/attacks.py:1830:80: E501 line too long (80 > 79 characters)
--------------------- >> end captured stdout << ----------------------

[nottombrown]

Lgtm. Thanks @juesato !