# -*- coding: utf-8 -*-
from accelbrainbase.computableloss._mxnet.l2_norm_loss import L2NormLoss
from accelbrainbase.extractabledata._mxnet.image_extractor import ImageExtractor
from accelbrainbase.iteratabledata._mxnet.unlabeled_image_iterator import UnlabeledImageIterator
from accelbrainbase.noiseabledata._mxnet.gauss_noise import GaussNoise
from accelbrainbase.observabledata._mxnet.convolutional_neural_networks import ConvolutionalNeuralNetworks
from accelbrainbase.observabledata._mxnet.neural_networks import NeuralNetworks
from accelbrainbase.observabledata._mxnet.convolutionalneuralnetworks.convolutional_auto_encoder import ConvolutionalAutoEncoder
from accelbrainbase.observabledata._mxnet.adversarialmodel.discriminative_model import DiscriminativeModel
from accelbrainbase.observabledata._mxnet.adversarialmodel.generative_model import GenerativeModel
from accelbrainbase.computableloss._mxnet.generator_loss import GeneratorLoss
from accelbrainbase.computableloss._mxnet.discriminator_loss import DiscriminatorLoss
from accelbrainbase.samplabledata.true_sampler import TrueSampler
from accelbrainbase.samplabledata.condition_sampler import ConditionSampler
from accelbrainbase.samplabledata.noisesampler._mxnet.uniform_noise_sampler import UniformNoiseSampler
from accelbrainbase.controllablemodel._mxnet.gan_controller import GANController
import mxnet as mx
import mxnet.ndarray as nd
import numpy as np
import pandas as pd
from mxnet.gluon.nn import Conv2D
from mxnet.gluon.nn import Conv2DTranspose
from mxnet.gluon.nn import BatchNorm
[docs]class GANImageGenerator(object):
'''
Image generation by GANs.
The Generative Adversarial Networks(GANs) (Goodfellow et al., 2014) framework establishes
a min-max adversarial game between two neural networks – a generative model, `G`, and a
discriminative model, `D`. The discriminator model, `D(x)`, is a neural network that computes
the probability that a observed data point `x` in data space is a sample from the data
distribution (positive samples) that we are trying to model, rather than a sample from our
generative model (negative samples).
Concurrently, the generator uses a function `G(z)` that maps samples `z` from the prior `p(z)` to
the data space. `G(z)` is trained to maximally confuse the discriminator into believing that
samples it generates come from the data distribution. The generator is trained by leveraging
the gradient of `D(x)` w.r.t. x, and using that to modify its parameters.
The Conditional GANs (or cGANs) is a simple extension of the basic GAN model which allows
the model to condition on external information. This makes it possible to engage the learned
generative model in different "modes" by providing it with different contextual
information (Gauthier, J. 2014).
This model can be constructed by simply feeding the data, `y`, to condition on to both the generator
and discriminator. In an unconditioned generative model, because the maps samples `z` from the prior
`p(z)` are drawn from uniform or normal distribution, there is no control on modes of the data being
generated. On the other hand, it is possible to direct the data generation process by conditioning
the model on additional information (Mirza, M., & Osindero, S. 2014).
References:
- Gauthier, J. (2014). Conditional generative adversarial nets for convolutional face generation. Class Project for Stanford CS231N: Convolutional Neural Networks for Visual Recognition, Winter semester, 2014(5), 2.
- Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., ... & Bengio, Y. (2014). Generative adversarial nets. In Advances in neural information processing systems (pp. 2672-2680).
- Makhzani, A., Shlens, J., Jaitly, N., Goodfellow, I., & Frey, B. (2015). Adversarial autoencoders. arXiv preprint arXiv:1511.05644.
- Mirza, M., & Osindero, S. (2014). Conditional generative adversarial nets. arXiv preprint arXiv:1411.1784.
- Salimans, T., Goodfellow, I., Zaremba, W., Cheung, V., Radford, A., & Chen, X. (2016). Improved techniques for training gans. In Advances in neural information processing systems (pp. 2234-2242).
- Zhao, J., Mathieu, M., & LeCun, Y. (2016). Energy-based generative adversarial network. arXiv preprint arXiv:1609.03126.
- Warde-Farley, D., & Bengio, Y. (2016). Improving generative adversarial networks with denoising feature matching.
'''
def __init__(
self,
dir_list,
width=28,
height=28,
channel=1,
initializer=None,
batch_size=40,
learning_rate=1e-03,
ctx=mx.gpu(),
discriminative_model=None,
generative_model=None,
):
'''
Init.
If you are not satisfied with this simple default setting,
delegate `discriminative_model` and `generative_model` designed by yourself.
Args:
dir_list: `list` of `str` of path to image files.
width: `int` of image width.
height: `int` of image height.
channel: `int` of image channel.
initializer: is-a `mxnet.initializer` for parameters of model.
If `None`, it is drawing from the Xavier distribution.
batch_size: `int` of batch size.
learning_rate: `float` of learning rate.
ctx: `mx.gpu()` or `mx.cpu()`.
discriminative_model: is-a `accelbrainbase.observabledata._mxnet.adversarialmodel.discriminative_model.DiscriminativeModel`.
generative_model: is-a `accelbrainbase.observabledata._mxnet.adversarialmodel.generative_model.GenerativeModel`.
'''
image_extractor = ImageExtractor(
width=width,
height=height,
channel=channel,
ctx=ctx
)
unlabeled_image_iterator = UnlabeledImageIterator(
image_extractor=image_extractor,
dir_list=dir_list,
batch_size=batch_size,
norm_mode="z_score",
scale=1.0,
noiseable_data=GaussNoise(sigma=1e-03, mu=0.0),
)
true_sampler = TrueSampler()
true_sampler.iteratorable_data = unlabeled_image_iterator
condition_sampler = ConditionSampler()
condition_sampler.true_sampler = true_sampler
computable_loss = L2NormLoss()
if discriminative_model is None:
output_nn = NeuralNetworks(
computable_loss=computable_loss,
initializer=initializer,
learning_rate=learning_rate,
learning_attenuate_rate=1.0,
attenuate_epoch=50,
units_list=[100, 1],
dropout_rate_list=[0.5, 0.0],
optimizer_name="SGD",
activation_list=["relu", "sigmoid"],
hidden_batch_norm_list=[BatchNorm(), None],
ctx=ctx,
hybridize_flag=True,
regularizatable_data_list=[],
scale=1.0,
output_no_bias_flag=True,
all_no_bias_flag=True,
not_init_flag=False,
)
d_model = ConvolutionalNeuralNetworks(
computable_loss=computable_loss,
initializer=initializer,
learning_rate=learning_rate,
learning_attenuate_rate=1.0,
attenuate_epoch=50,
hidden_units_list=[
Conv2D(
channels=16,
kernel_size=6,
strides=(2, 2),
padding=(1, 1),
),
Conv2D(
channels=32,
kernel_size=3,
strides=(2, 2),
padding=(1, 1),
),
],
input_nn=None,
input_result_height=None,
input_result_width=None,
input_result_channel=None,
output_nn=output_nn,
hidden_dropout_rate_list=[0.5, 0.5],
hidden_batch_norm_list=[BatchNorm(), BatchNorm()],
optimizer_name="SGD",
hidden_activation_list=["relu", "relu"],
hidden_residual_flag=False,
hidden_dense_flag=False,
dense_axis=1,
ctx=ctx,
hybridize_flag=True,
regularizatable_data_list=[],
scale=1.0,
)
discriminative_model = DiscriminativeModel(
model=d_model,
initializer=None,
learning_rate=learning_rate,
optimizer_name="SGD",
hybridize_flag=True,
scale=1.0,
ctx=ctx,
)
else:
if isinstance(discriminative_model, DiscriminativeModel) is False:
raise TypeError("The type of `discriminative_model` must be `DiscriminativeModel`.")
if generative_model is None:
g_model = ConvolutionalNeuralNetworks(
computable_loss=computable_loss,
initializer=initializer,
learning_rate=learning_rate,
learning_attenuate_rate=1.0,
attenuate_epoch=50,
hidden_units_list=[
Conv2DTranspose(
channels=16,
kernel_size=6,
strides=(1, 1),
padding=(1, 1),
),
Conv2DTranspose(
channels=1,
kernel_size=3,
strides=(1, 1),
padding=(1, 1),
),
],
input_nn=None,
input_result_height=None,
input_result_width=None,
input_result_channel=None,
output_nn=None,
hidden_dropout_rate_list=[0.5, 0.0],
hidden_batch_norm_list=[BatchNorm(), None],
optimizer_name="SGD",
hidden_activation_list=["relu", "identity"],
hidden_residual_flag=False,
hidden_dense_flag=False,
dense_axis=1,
ctx=ctx,
hybridize_flag=True,
regularizatable_data_list=[],
scale=1.0,
)
generative_model = GenerativeModel(
noise_sampler=UniformNoiseSampler(
low=-1e-05,
high=1e-05,
batch_size=batch_size,
seq_len=0,
channel=channel,
height=height,
width=width,
ctx=ctx
),
model=g_model,
initializer=None,
condition_sampler=condition_sampler,
conditonal_dim=1,
learning_rate=learning_rate,
optimizer_name="SGD",
hybridize_flag=True,
scale=1.0,
ctx=ctx,
)
else:
if isinstance(generative_model, GenerativeModel) is False:
raise TypeError("The type of `generative_model` must be `GenerativeModel`.")
GAN = GANController(
true_sampler=true_sampler,
generative_model=generative_model,
discriminative_model=discriminative_model,
generator_loss=GeneratorLoss(weight=1.0),
discriminator_loss=DiscriminatorLoss(weight=1.0),
feature_matching_loss=L2NormLoss(weight=1.0),
optimizer_name="SGD",
learning_rate=learning_rate,
learning_attenuate_rate=1.0,
attenuate_epoch=50,
hybridize_flag=True,
scale=1.0,
ctx=ctx,
initializer=initializer,
)
self.GAN = GAN
[docs] def learn(self, iter_n=1000, k_step=10):
'''
Learning.
Args:
iter_n: `int` of the number of training iterations.
k_step: `int` of the number of learning of the `discriminative_model`.
'''
self.GAN.learn(
iter_n=iter_n,
k_step=k_step,
)