Released Neural Network Libraries v.1.0.15! Added Spectral Normalization and 3 New Models (VGG, DenseNet, NIN)!

Friday, April 05, 2019


Posted by shin

We have released Neural Network Libraries v.1.0.15! Added spectral normalization and 3 new models (VGG, DenseNet, NIN)!


Spectral Normalization

Spectral normalization (https://arxiv.org/abs/1802.05957) is a novel weight normalization technique to stabilize the training of the discriminator that is computationally light.

import nnabla as nn
import nnabla.parametric_functions as PF

b, c, h, w = 4, 64, 32, 32

# Spectrally normalized convolution
apply_w = lambda w: PF.spectral_norm(w, dim=0)
h = nn.Variable.from_numpy_array(np.random.randn(b, c, h, w))
h = PF.convolution(h, with_bias=False, apply_w=apply_w)


New NNabla Models: VGG,DenseNet,Network in Network

3 new models have been added to NNabla Models and can be used with pre-trained parameters!

VGG (https://arxiv.org/abs/1409.1556)

DenseNet (https://arxiv.org/abs/1608.06993)

Network in Network (https://arxiv.org/abs/1312.4400)

Other Updates

Function Layers




Bug Fix