RevNets improve generative performance on CelebA dataset.
problem Training autoencoders is complicated by needing separate encoder and decoder models.
method Used adversarial autoencoder framework with RevNets in latent space.
result RevNets generate coherent faces with similar quality to Variational Autoencoders.
Fairness GAN generates fair datasets for decision making.
problem Creating fair datasets for allocative decision making.
method Novel auxiliary classifier GAN aiming for demographic parity or equality of opportunity.
result Improves demographic parity and equality of opportunity in generated images.
EvalGAN evaluates GANs by measuring reconstruction quality and likelihood.
problem Evaluating the quality of generated images from GANs.
method EvalGAN uses a test set to measure reconstruction quality and likelihood in the original sample space.
result EvalGAN provides a direct and agnostic method for evaluating GANs.
Soft Truncation improves diffusion model performance by balancing loss scales across diffusion times.
problem Inverse correlation between density estimation and sample generation in diffusion models.
method Introduces Soft Truncation, a training technique that softens the truncation hyperparameter.
result Soft Truncation achieves state-of-the-art performance on various datasets.
Proposes a potential flow generator for generative models.
problem Improving the correctness and robustness of generative models.
method Integrates L2 optimal transport regularity into generative models. result Demonstrates effectiveness in image translation tasks.
Hierarchical model learns progressively abstract representations.
problem Learning meaningful hierarchical latent structures.
method Adversarial training for generative and inference models.
result Minimizing Jensen-Shanon divergence minimizes reconstruction error.
SVD-RND detects blurred images better than conventional methods.
problem Blurred images can fool conventional OOD detection schemes.
method Constructs a novel RND-based detector that uses blurred images during training.
result SVD-RND outperforms baseline detectors in various domains.
Improved GANs produce high-quality images with stable training and increased variation.
problem Training GANs for high-quality images with stability and diversity.
method Progressive growth of generator and discriminator, adding layers progressively.
result Achieved unprecedented image quality and record inception score.
Probabilistic inpainting learns multiple plausible images from missing data.
problem Generating multiple plausible images from missing data in images.
method Building a PixelCNN model that learns a distribution of images conditioned on visible pixels.
result The method produces diverse and realistic inpaintings.
Compares GAN and VAE models, highlighting differences in fidelity and mode collapse.
problem Understanding and comparing GAN and VAE models.
method Empirical comparison of GAN and VAE models on MNIST and CelebA datasets.
result Hypothesis explaining differences in fidelity and mode collapse between GAN and VAE models.
NVAE improves VAE performance on large image datasets.
problem Improving variational autoencoder performance for large image datasets.
method Deep hierarchical VAE with depth-wise separable convolutions and batch normalization, residual parameterization of Normal distributions, and spectral regularization.
result NVAE achieves state-of-the-art results on MNIST, CIFAR-10, CelebA 64, and CelebA HQ datasets.
Deep AA generates latent archetypes from datasets.
problem Representing high-dimensional datasets in understandable basic entities.
method Extends linear Archetypal Analysis with deep learning capabilities.
result Reduces dependence on expert knowledge and handles side information.
GANchors generates realistic image perturbations for better classifier explanations.
problem Improving the trustworthiness of image classification explanations.
method Using GANs to optimize a lower-dimensional latent space for realistic perturbation distributions.
result Generated images are more likely to be from the original training set, leading to more precise explanations.
Study on shortcuts in deep networks, revealing their layer-wise distribution and impact.
problem Understudied impact of shortcuts on feature representations in deep networks.
method Layer-wise localization through counterfactual training on clean and skewed datasets.
result Shortcuts are distributed throughout the network, not localized in specific layers.
Generates samples from a target distribution using a discriminator.
problem Sampling from complex target distributions.
method Implicit Metropolis-Hastings algorithm using GAN discriminator.
result Discriminator loss bounds the distance to target distribution.
Paper tackles co-generation with GANs, developing a new algorithm.
problem Inferring the most likely configuration for a subset of variables given the rest.
method Annealed Importance Sampling based Hamiltonian Monte Carlo algorithm.
result Significantly outperforms classical gradient based methods.
MetalGAN synthesizes images across multiple domains without labels.
problem Synthesizing images across multiple domains using a single network.
method Combines cGAN for image generation and Meta-Learning for domain switch.
result MetalGAN successfully produces multi-domain images without hard-coded labels.
IntroVAC learns interpretable latent subspaces for better image quality.
problem Difficulties in interpreting latent spaces and limitations in image generation.
method Introspective Variational Classifier (IntroVAC) using additional labels and adversarial training.
result Improved image quality and meaningful latent directions for fine-grained manipulation.
Improves GAN training with a repulsive loss function.
problem Discourages learning of fine details in data.
method Proposes a repulsive loss function and a bounded Gaussian kernel.
result Significantly improves GAN performance without additional computational cost.
New method learns sparse distributions by thresholding samples, improving performance and efficiency.
problem Sparse coding optimization in high-dimensional problems is computationally expensive and inefficient.
method Proposes a new variational sparse coding approach that learns sparse distributions by thresholding samples.
result Shows superior performance, statistical efficiency, and gradient estimation compared to other sparse distributions.
SeqRF straightens generative model flows to speed up sampling.
problem High global truncation error in ODE-based solvers for generative models.
method SeqRF, a learning technique that straightens the probability flow.
result Significantly improved sampling speed and synthesis quality.
WS diffusion models handle anisotropic Gaussian noise better than conventional methods.
problem Handling anisotropic Gaussian noise in imaging inverse problems.
method Whitened Score (WS) diffusion models based on stochastic differential equations.
result WS DMs outperform conventional DMs on anisotropic Gaussian noise.
EMA outperforms MA in GAN training, reducing cycle amplitudes and improving stability.
problem Improving GAN training stability and effectiveness.
method Comparison of Moving Average (MA) and Exponential Moving Average (EMA) techniques.
result EMA converges to limit cycles with vanishing amplitude in simple bilinear games and enhances GAN training stability.
New method learns identity-preserving transformations on data manifolds without labels.
problem Learning identity-preserving transformations on natural variations without supervision.
method Introduces a learning strategy that does not require transformation labels and learns local regions for operators.
result Trains on MNIST and Fashion MNIST, and CelebA, learning transformations without labels.
CausalVAE learns causal relationships in VAE models for better data disentanglement.
problem Learning disentanglement of independent factors from observational data.
method CausalVAE framework with a Causal Layer to transform exogenous factors into causal endogenous ones.
result CausalVAE learns semantically interpretable causal representations and accurately identifies their DAG structure.
Faster reconstruction of compressed signals using conditional GAN and NPGD.
problem Recovering compressed signals from measurements.
method Network-based projected gradient descent (NPGD) combined with measurement-conditional generative adversarial networks (GANs/BEGANs).
result Significant speed-up in reconstruction (up to 140-175 times faster).
Improves GAN performance without changing network architecture.
problem Discriminator fails to guide generator effectively.
method Cascading Rejection (CR) module for discriminator.
result Significantly improves GAN performance in terms of FID.
Venn GAN models multiple distributions, discovering shared and unique aspects.
problem Modeling and understanding multiple data distributions effectively.
method A GAN design with shared and non-shared generator distributions.
result Effective modeling of various datasets (MNIST, Fashion MNIST, CIFAR-10, Omniglot, CelebA).
GILBO measures complexity of latent variable models.
problem Measuring the complexity of latent variable models.
method Proposes GILBO, a lower bound on mutual information, to measure model complexity.
result GILBO provides a data-independent measure of model complexity.
VAE fails to encode typical samples, new method improves robustness.
problem VAE does not necessarily encode typical samples generated by its decoder.
method Alternative construction of variational approximation distribution, self-consistency approach.
result Encoders trained with self-consistency approach are robust to adversarial attacks.
Paper tackles unsupervised learning of 3D shapes from single images.
problem Learning 3D shapes from single images without supervision.
method Generative models, variational auto-encoders, adversarial methods.
result Model learns 3D shapes and poses from single images, showing potential for various datasets.
New method for visualizing high-level concepts in generative models.
problem Challenges in evaluating and visualizing concepts in generative models.
method Introduces a method to compute concept saliency maps for latent representations of known or novel high-level concepts.
result Concept saliency maps highlight input features important for high-level concepts.
LSGM trains SGMs in latent space for faster sampling.
problem Efficiently generating high-quality samples from complex distributions.
method LSGM trains SGMs in latent space using a variational autoencoder framework, introducing new score-matching objectives and parameterizations.
result LSGM achieves state-of-the-art FID score of 2.10 on CIFAR-10 and outperforms previous SGMs in sampling time.
Paper presents a new algorithm to approximate Wasserstein-2 barycenters without bias.
problem Approximating Wasserstein-2 barycenters of continuous measures.
method Generative model approach using arbitrary neural networks.
result The method does not introduce bias and is applicable to large-scale tasks.
Generative adversarial network synthesizes sketches into realistic images.
problem Improving the quality and realism of facial sketches.
method Hybrid GAN with quality guided encoder and identity preserving network.
result Synthesized images are more realistic and maintain identity.
Generates realistic faces from detailed textual descriptions.
problem Face generation from fine-grained textual descriptions.
method Conditional GAN model with DC-GAN and GAN-CLS loss, using CelebA dataset with generated captions.
result Promising results in generating diverse face images from fine-grained textual descriptions.
Improved VAEs by training a contrastive prior to match posterior.
problem Prior hole problem in VAEs, leading to poor image generation.
method Introduced a contrastive energy-based prior and trained it using noise contrastive estimation.
result Significant improvement in VAE generative performance on various datasets.
This paper describes InfoGAN, an information-theoretic extension to the Generative Adversarial Network that is able to learn disentangled representations in a completely unsupervised manner. InfoGAN is a generative adversarial network that also maximizes the mutual information between a small subset of the latent varia…
Proposes a technique to interpret deep learning models by generating counterfactual inputs.
problem Understanding and explaining the decisions made by deep neural networks.
method Uses a generative model to edit input images and generate counterfactual scenarios for model interpretation.
result Demonstrates the effectiveness of the introspection approach on MNIST and CelebA datasets.
SepVAE separates patient-specific patterns from healthy ones using contrastive VAE.
problem Separating patient-specific patterns from healthy ones in medical datasets.
method SepVAE uses a contrastive VAE with disentangling and classification losses to differentiate between common and salient features.
result SepVAE outperforms previous methods in three medical applications and a CelebA dataset.
CSVAE learns interpretable latent subspaces for binary labels.
problem Learning interpretable latent representations correlated to specific labels.
method Conditional Subspace VAE (CSVAE) using mutual information minimization.
result CSVAE extracts interpretable latent subspaces for binary labels.
Bidirectional VAE reduces parameters and improves image tasks.
problem Improving image reconstruction, classification, interpolation, and generation.
method Uses a single neural network for both encoding and decoding in both forward and backward directions.
result Bidirectional VAEs reduce parameters by almost 50% and slightly outperform unidirectional VAEs.
Efficiently approximates fairness-accuracy trade-offs for diverse datasets.
problem Inherent trade-off between fairness and accuracy in machine learning models.
method You-Only-Train-Once (YOTO) framework for computationally efficient approximation.
result Robust methodology for auditing model fairness with statistical guarantees.
Improves training GANs by escaping limit cycles.
problem Limit cycling behavior in training GANs.
method Predictive Centripetal Acceleration Algorithm (PCAA) combined with Adam.
result PCAA improves convergence rates and effectively trains GANs.
Improved score-based models generate high-quality images up to 256x256.
problem Training score-based models for high-resolution images is unstable and limited.
method Theoretical analysis, exponential moving average of model weights.
result Score-based models can generate high-fidelity images up to 256x256.
DAEs can generate images without additional loss terms, inheriting VAE properties.
problem Difficulty in using VAEs for practical generative modelling.
method Empirical exploration of DAEs for image generation without novel methods.
result DAEs can generate images successfully without additional loss terms.
FedGAN trains GANs across distributed data sources with reduced communication.
problem Training GANs across non-independent data sources with privacy and communication constraints.
method FedGAN uses local generators and discriminators synced via an intermediary, proving convergence under standard assumptions.
result FedGAN converges and performs similarly to general distributed GANs with reduced communication complexity.
DALI improves inference for GANs by matching prior and conditional distributions.
problem Effective inference for generative adversarial models (GANs).
method Explicitly matches prior and conditional distributions in data and code spaces, constraining dependency structure.
result Significantly improves reconstruction and generation compared to other GAN inference models.