behave in the way you would expect.

with respect to parameters and one that updates the parameters with those
gradients.

- Made layer() recurse into repeat objects so that the index given to layer()
  does what you would expect.
- Added an operator<< for network objects that prints the network architecture.

thing to do for this kind of training.

--HG--
rename : examples/dnn_mnist_resnet_ex.cpp => examples/dnn_mnist_advanced_ex.cpp

Added simd8 usage on FHOG features computation

running variance output.

to automatically select the right mode.  The serialization format for bn_
layers has also changed, but the code will still be able to deserialize older
bn_ objects.

realistic.