Index

Search in book...
Toggle Font Controls
Create new playlist

Name your new playlist

Playlist description (optional)
Sign In

Email address

Password

Forgot Password?

or

Continue with Facebook

Continue with Google
Sign Up

Full Name

Email address

Confirm Email Address

Password

or

Continue with Facebook

Continue with Google

Previous Chapter

Training Generative Adversarial Networks (GANs)

List of Figures

Index

[A][B][C][D][E][F][G][H][I][J][K][L][M][N][O][P][R][S][T][U][V][W][Z]

A

accuracy testing ACGAN activation maps Adam algorithm, 2^nd, 3^rd adaptive moment estimation adversarial attacks adversarial examples adversaries to GANs context of distributions noise security implications of signal training adversarial loss adversarial training, 2^nd cost functions training process AI ethics AlphaGo artificial general intelligence autoencoders GANs and generation with high-level function of overview of using variational

B

backpropagated errors back-translation batch normalization. See BN (batch normalization). BEGAN BigGAN binary cross-entropy, 2^nd binary_crossentropy function BN (batch normalization), 2^nd, 3^rd, 4^th, 5^th, 6^th computing overview of

C

CapsuleNets categorical_crossentropy function CGAN (Conditional GAN), 2^nd, 3^rd, 4^th, 5^th, 6^th architecture diagram building models Discriminator generating targeted data Generator, 2^nd implementing outputting sample images overview of setup training training models CGI (computer-generated imagery) classification classification loss classifiers, fully supervised Cloud AutoML, Google CNNs (convolutional neural networks), 2^nd CNTK (Cognitive Toolkit), Microsoft colorization computer vision (CV) computer-generated imagery (CGI) Conda, 2^nd Conditional GAN. See CGAN. confusion matrix contraction path conv layer conv2d() function Conv2DTranspose convenience function, Keras ConvNet (convolutional neural network) convolutional filters parameter sharing visualizing ConvNets, 2^nd convolutional filters convolutional kernel core Discriminator network cost functions covariate shift cross-entropy loss CV (computer vision), 2^nd CyCADA (Cycle Consistent Adversarial Domain Adaptation) cycle-consistency loss, 2^nd CycleGAN, 2^nd, 3^rd, 4^th, 5^th adversarial loss applications of architecture of building networks Discriminator architecture Generator architecture augmentations building networks cycle-consistency loss expansions identity loss image-to-image translation object-oriented design of GANs running

D

data denoising data, targeted Dataset object datasets DAWNBench DCGAN (Deep Convolutional GAN), 2^nd, 3^rd, 4^th, 5^th batch normalization computing overview of building ConvNet convolutional filters parameter sharing visualizing generating handwritten digits with implementing Discriminator implementing Generator importing modules model output specifying model input dimensions history of running decision trees (DT) decoder network, 2^nd, 3^rd, 4^th, 5^th deconv layer deconv2d() function deconvolutional layers, 2^nd Deep Convolutional GAN. See DCGAN. deep learning algorithms Deep Learning with Python (Chollet), 2^nd deep neural networks (DNNs), 2^nd DeepFakes DeepMind DeOldify design of GANs Discriminator architecture of building confusion matrix core network Generator vs. implementing, 2^nd supervised training unsupervised Discriminator network, 2^nd, 3^rd, 4^th, 5^th, 6^th, 7^th, 8^th, 9^th, 10^th, 11^th, 12^th, 13^th, 14^th, 15^th, 16^th, 17^th discriminator.trainable setting distributions DNNs (deep neural networks), 2^nd Docker Hub domain transferred image don’t repeat yourself (DRY) dropout DRY (don’t repeat yourself) DT (decision trees) dynamic range

E

ϵ (epsilon) earth mover’s distance, 2^nd, 3^rd Embedding layers, 2^nd encoder network, 2^nd, 3^rd, 4^th, 5^th ensembles (Ens.) equalized learning rates ethics expanding path expectation (Es) explicit loss function explicit objective function, 2^nd

F

fake images fake labels false negative false positive fast sign gradient method (FSGM) fc (fully connected) layer feature detectors FFDM (full-field digital mammography) FID (Fréchet inception distance), 2^nd, 3^rd, 4^th, 5^th filters. See convolutional filters. fit function foolbox class Fréchet inception distance. See FID. FSGM (fast sign gradient method) full-field digital mammography (FFDM) fully connected (fc) layer fully supervised classifiers functional API, 2^nd

G

game setups, 2^nd GAN Zoo Ganbreeder app GANs (Generative Adversarial Networks) adversaries to autoencoders and in action in fashion in medicine, 2^nd Nash equilibrium object-oriented design of overview of practical application of training why study Gaussian, 2^nd, 3^rd, 4^th, 5^th GDPR (General Data Protection Regulation) generation with autoencoders Generative Adversarial Networks. See GANs. generative modeling Generator network, 2^nd, 3^rd, 4^th, 5^th, 6^th, 7^th, 8^th, 9^th, 10^th, 11^th, 12^th, 13^th, 14^th, 15^th, 16^th, 17^th confusion matrix Discriminator vs. pixel-wise feature normalization in global variables Goodfellow, Ian adversarial examples invention of GANs label smoothing recommendation SVHN benchmark Google Earth GP (gradient penalties) gradient ascent gradient descent gradient-descent-based optimizer gradients penalizing sparse growing layers

H

handwritten digits adversarial training cost functions training process building models Discriminator confusion matrix Generator vs. implementing generating, 2^nd Generator confusion matrix Discriminator vs. implementing importing statements inspecting results outputting sample images running models training algorithms, 2^nd hyperparameters, 2^nd, 3^rd hyperrealistic imagery

I

ICLR (International Conference on Learning Representations), 2^nd identity loss ImageNet dataset, 2^nd, 3^rd, 4^th, 5^th images, outputting, 2^nd image-to-image translation, 2^nd imitation game importing modules statements Inception network, 2^nd inception scores, 2^nd Inception V3 network, 2^nd inputs, normalizing InstanceNormalization interclass International Conference on Learning Representations (ICLR), 2^nd interpolating latent space intraclass, 2^nd IS (inception score), 2^nd Iterative training/tuning

J

JSD (Jensen-Shannon divergence), 2^nd, 3^rd

K

Keras library, 2^nd Keras_contrib keras.datasets Keras-GAN repository, 2^nd keras.layers.BatchNormalization function KL (Kullback–Leibler) divergence, 2^nd, 3^rd, 4^th kNN (nearest neighbors)

L

labels lambda_cycle hyperparameter lambda_id hyperparameter LAPGAN Laplacian pyramid latent space, 2^nd, 3^rd, 4^th, 5^th latent space interpolation latent vector latent_vector layers Leaky ReLU activation function, 2^nd, 3^rd learnable classification functions learning rates loss function, 2^nd, 3^rd LR (logistic regression)

M

machine learning (ML) MAE (mean absolute error) MammoGAN maximum likelihood estimation, 2^nd MaxPool mean squared deviation mean squared error (MSE) mini-batch standard deviation ML (machine learning) MM-GAN (Min-Max GAN) MNIST dataset, 2^nd, 3^rd, 4^th, 5^th, 6^th, 7^th, 8^th, 9^th mode collapse, 2^nd modules, importing MSE (mean squared error) Multiply layer

N

Nash equilibrium, 2^nd nearest neighbors (kNN) network depth networks, 2^nd, 3^rd neural networks NeurIPS, formerly NIPS (Neural Information Processing Systems) NLP (natural language processing) noise noisy labels Non-Saturating GAN. See NS-GAN. normalizing features in Generator inputs np.random.normal NS-GAN (Non-Saturating GAN), 2^nd, 3^rd, 4^th num_labeled parameter

O

object recognition algorithms object-oriented design OCR (optical character recognition) off-by-one cycle one-class classifier one-hot-encoded labels overall loss overgeneralization

P

parameters, sharing PatchGAN architecture PCA (principal component analysis) penalizing gradients PGD (projected gradient descent) PGGAN (Progressive GAN), 2^nd pix2pix, 2^nd point estimates predict() method, 2^nd prediction preference maximization principal component analysis (PCA) ProGAN (Progressive GAN), 2^nd, 3^rd, 4^th, 5^th progressive growing projected gradient descent (PGD) PyTorch Hub

R

random noise vectors, 2^nd reconstruction loss, 2^nd regression regular convolution relative entropy ReLU (rectified linear unit) activation ResNet-50, 2^nd, 3^rd reward function RGAN (Relativistic GAN) RMSE (root mean squared error) RMSprop, 2^nd Robust Manifold Defense ROC curves, 2^nd rotations

S

saccades SAGAN (Self-Attention GAN), 2^nd sample images, outputting, 2^nd sample_images() function, 2^nd sample_interval iterations saturation scaling Self-Attention GAN. See SAGAN. self-supervised machine learning set_title()method SGAN (Semi-Supervised GAN), 2^nd, 3^rd, 4^th architecture diagram building models datasets Discriminator core network supervised unsupervised fully supervised classifiers vs. Generator implementing overview of architecture training objectives training process setup training testing accuracy training models SGD (stochastic gradient descent), 2^nd, 3^rd sharing parameters sigmoid activation function, 2^nd, 3^rd signal skip connections sliced Wasserstein distance (SWD), 2^nd slow convergence smoothing layers soft labels softmax function SPADE (GauGAN) sparse gradients spectral normalization st (strides) standard deviations statements, importing stochastic gradient descent (SGD), 2^nd, 3^rd Style GAN supervised classifiers supervised Discriminator supervised loss SVHN (Street View House Numbers) dataset SVM (support-vector machine) SWD (sliced Wasserstein distance), 2^nd

T

tanh activation function, 2^nd, 3^rd, 4^th targeted data TensorFlow Extended TensorFlow framework, 2^nd, 3^rd, 4^th testing accuracy TFHub (TensorFlow Hub), 2^nd, 3^rd tf.shape list Theano framework Towards Data Science training algorithms, 2^nd Training dataset translation, image-to-image transposed convolutions, 2^nd, 3^rd true distributions, 2^nd true negative true positive truncation tricks Turing test

U

U-Net architecture unsupervised learning autoencoders generation using autoencoders variational autoencoders unsupervised loss

V

VAE (variational autoencoder), 2^nd, 3^rd variational autoencoders VGG-19 trained network vid2vid

W

Wasserstein distance WGAN (Wasserstein GAN), 2^nd, 3^rd, 4^th, 5^th WGAN-GP (gradient penalty) version, 2^nd

Z

z_dim variable zero-sum games

..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.