Gan

Ok, so I think that the modifications I made here can be merged ...

Two main changes as compared to the master:

1. Added some layers
2. Modified the way variable name and scope is handled in layers (and in SequenceNetwork as well)

Change 1 should by no means affect what has been done until now (on the master). However, change 2 may require a bit of adaptation, especially when defining architectures and layers ...

@tiago.pereira I'll let you check and then we can discuss

Also, I think that this will be irrelevant when we'll go to tf-slim anyway !

bob/learn/tensorflow/datashuffler/Noise.py

+#!/usr/bin/env python
+# encoding: utf-8
+# Guillaume HEUSCH <guillaume.heusch@idiap.ch>
+# Mon  8 May 09:30:15 CEST 2017
+
+import numpy
+import tensorflow as tf
+
+
+class Noise(object):
+
+    """
+    This datashuffler implements a generator of noise to feed a GAN's generator/ 
+
+     **Parameters**
+
+     input_shape:
+       The shape of the inputs
+
+     input_dtype:
+       The type of the data,
+
+     batch_size:
+       Batch size
+
+     seed:
+       The seed of the random number generator
+    """
+
+    def __init__(self,
+                 input_shape,
+                 input_dtype="float64",
+                 batch_size=1,
+                 seed=10,
+                 ):
+
+        self.input_shape = tuple(input_shape)
+        self.shape = tuple([batch_size] + input_shape)
+        self.input_dtype = input_dtype
+        self.batch_size = batch_size
+
+        # Setting the seed
+        numpy.random.seed(seed)
+
+        self.data_placeholder = None
+        self.label_placeholder = None
+
+    def make_data(self):
+      self.data = np.random.uniform(-1., 1., size=[self.batch_size, self.input_shape])
+
+    def get_placeholders(self, name=""):
+      if self.data_placeholder is None:
+        self.data_placeholder = tf.placeholder(tf.float32, shape=self.shape, name=name)
+
+      if self.label_placeholder is None:
+        self.label_placeholder = tf.placeholder(tf.int64, shape=self.shape[0])
+
+      return self.data_placeholder
+
+    def get_batch(self):
+        """
+        Get a random batch of noise input.
+        The noise is drawn from a uniform distribution in interval [-1 1]
+
+        ** Returns **
+
+        data:
+          Selected samples
+
+        """
+        noise = numpy.random.uniform(-1., 1., size=[self.batch_size, self.input_shape[0]])
+        
+        return noise