Unified framework for learning function representations using INRs and Transformers.
problem Scalability and efficiency limitations in existing generative models.
method Integrates INRs and Transformer-based hypernetworks into latent variable models.
result Improved scalability, expressiveness, and generalization over existing models.
Study on DiTs' rates of approximation and estimation under various data assumptions.
problem Investigating statistical rates of conditional diffusion transformers.
method Discretization and Taylor expansion of conditional diffusion score function under Hölder smooth data assumption.
result Establishes statistical limits for conditional and unconditional DiTs, offering practical guidance.
Latent DiTs improve data distribution recovery and inference efficiency under low-dimensional latent space.
problem Improving data distribution recovery and inference efficiency in latent DiTs.
method Investigates statistical and computational limits of latent DiTs under low-dimensional latent space assumption, deriving approximation error bounds, sample complexity, and efficient inference and training algorithms.
result Latent DiTs can bypass high dimensionality challenges and achieve almost-linear time inference and training.
New model generates realistic single-cell gene expression data.
problem Generating realistic single-cell gene expression profiles is challenging.
method scLDM, a latent diffusion model using Diffusion Transformers and linear interpolants.
result Superior performance in generating realistic single-cell gene expression data.
In deep latent Gaussian models, the latent variable is generated by a time-inhomogeneous Markov chain, where at each time step we pass the current state through a parametric nonlinear map, such as a feedforward neural net, and add a small independent Gaussian perturbation. This work considers the diffusion limit of suc…
Improved image translation using asymmetric gradient guidance.
problem Trade-off between style transformation and content preservation in diffusion models.
method Asymmetric gradient guidance to guide reverse diffusion sampling.
result Our method outperforms state-of-the-art models in image translation tasks.
LADD models improve discrete diffusion for faster language generation.
problem Practical discrete diffusion models ignore cross-token dependencies, degrading performance.
method Introduces a learnable auxiliary latent channel, diffusing over the joint (token, latent) space.
result LADD models yield improvements on unconditional generation metrics.
Latent diffusion improves robustness in missing data imputation.
problem Missing data imputation under MCAR corruption.
method Two-stage framework: VAE for latent feature learning, diffusion model in latent space.
result Latent diffusion maintains high quality and stability up to 50% missingness.
LSI enables joint learning of latent variables and generative models.
problem Joint optimization of latent variables and generative models.
method Developed a principled ELBO objective in continuous time for joint learning.
result LSI learns effective latent representations and generative transformations.
Deep generative models are universal tools for learning data distributions on high dimensional data spaces via a mapping to lower dimensional latent spaces. We provide a study of latent space geometries and extend and build upon previous results on Riemannian metrics. We show how a class of heuristic measures gives mor…
Few-Shot Diffusion Models generate new samples from small image sets.
problem Generating new samples from a few images.
method Conditional Denoising Diffusion Probabilistic Models (DDPM) with patch-based input set information.
result FSDM can generate samples from previously unseen classes conditioned on as few as 5 samples.
Residual Prior Diffusion integrates coarse latent priors with diffusion models for better generative tasks.
problem Diffusion models struggle with representing both large-scale and fine-scale details in data distributions.
method Two-stage framework: first a coarse prior model captures large-scale structure, then a diffusion model represents the residual.
result RPD accurately captures fine-scale details while preserving large-scale structure, outperforming standard diffusion models.
This paper uses diffusion models for lossy image compression, improving perceptual metrics and practicality.
problem Lossy image compression with improved perceptual metrics and practicality.
method End-to-end optimized lossy image compression using conditional diffusion models.
result The model yields stronger FID scores and competitive performance in distortion metrics.
New method improves image generation for inverse problems using text prompts.
problem Suboptimal performance of existing latent diffusion models for inverse problems.
method Prompt tuning and latent variable projection to optimize text embeddings.
result P2L method outperforms existing methods on various inverse problem tasks.
Boomerang generates nonidentical images similar to input on image manifolds.
problem Generating nonidentical images similar to input on image manifolds.
method Adding noise to input image, moving closer to latent space, and mapping back through partial reverse diffusion.
result Boomerang generates nonidentical images similar to input on image manifolds.
Improved image generation quality using closed-form discriminator guidance in diffusion models.
problem Enhancing the quality of images generated by diffusion models.
method Theoretical framework to analyze GAN discriminator's effect on Langevin sampling, proposing IPM-GAN optimization as smoothed score-matching.
result Closed-form kernel-based discriminator guidance improves metrics like CLIP-FID and KID.
A new TTS method uses diffusion and VAE for better speech synthesis.
problem Improving text-to-speech synthesis for better speech quality and robustness.
method Combines diffusion probabilistic model and variational autoencoder for latent variable conversion.
result The method is robust to poor orthography and alignment errors.
We solve continuous-time latent SDE identifiability using diffusion shifts.
problem Identifiability of latent SDEs in continuous-time time series.
method Environment-induced shifts in diffusion covariance for additive-noise latent SDEs.
result Two diagonal diffusion regimes with distinct variance ratios identify latent coordinates up to permutation and scaling.
Combines deep state space models with diffusion models for better forecasting and capturing latent dynamics
problem Forecasting and capturing latent dynamics in time series
method DDSSM: Diffusion-driven state space model
result Empirically outperforms state-of-the-art deep SSM
New method uses diffusion models to solve inverse problems.
problem Solving ill-posed inverse problems with powerful priors.
method Formulate posterior sampling as a regularized Wasserstein gradient flow in latent space.
result Demonstrates improved performance on standard benchmarks.
DDVI uses diffusion models for variational inference, improving latent variable model performance.
problem Improving variational inference in latent variable models.
method Introduces diffusion-based variational posteriors trained with a regularized ELBO.
result Outperforms alternative variational posteriors on various benchmarks and a biology task.
Paper introduces a new generative learning model using Schrödinger bridge diffusion in latent space.
problem Learning distributions from divergent data distributions.
method Pre-training with large-scale models, Schrödinger bridge diffusion model in latent space.
result Effective control of second-order Wasserstein distance between generated and target distributions.
Proposes LDIDPs for efficient sequential data generation from latent dynamical models.
problem Challenges in generating high-fidelity sequential samples from latent dynamical models.
method Utilizes implicit diffusion processes to sample from latent dynamical processes.
result Demonstrates accurate learning of dynamics and efficient generation of high-quality sequential data.
Co-Diffusion predicts drug-target affinity by learning latent manifolds and diffusion, improving generalization.
problem Cold-start regimes in drug-target affinity prediction due to label scarcity and domain shifts.
method Two-stage framework: latent manifold alignment and latent diffusion regularization.
result Significantly outperforms state-of-the-art baselines, especially in zero-shot generalization.
Study on reliability of latent reuse in diffusion models under distribution shift.
problem When can latent spaces from a source dataset be reused for a target dataset with different distributions?
method Considered a source-target setting with approximately low-dimensional datasets near different subspaces. Analyzed the target-domain score error due to principal-angle misalignment and target ambient noise.
result Latent reuse is reliable only if the source and target subspaces are close and the target ambient noise is not too amplified.
This study rethinks the latent space in generative modeling, improving performance with less complex models.
problem Determining the optimal latent space for generative models and understanding its impact on model complexity.
method Proposed a new distance metric between latent and data distributions, and a two-stage training strategy called Decoupled Autoencoder (DAE).
result Improves generative performance with less complex models, as shown by comprehensive experiments on various models.
CADD improves generative quality by augmenting discrete diffusion with continuous latent space.
problem Loss of semantic information between denoising steps in discrete diffusion models.
method Introduces a framework that augments discrete state space with a continuous latent space, allowing for graded, informative masked tokens.
result CADD improves generative quality across text generation, image synthesis, and code modeling.
Paper analyzes latent space geometry in generative models using Fisher information.
problem Understanding the structure of latent spaces in generative models.
method Reconstructs Fisher information metric from generated samples and posterior distribution.
result Reveals fractal structure and abrupt changes in Fisher metric at phase boundaries.
New algorithm trains latent diffusion models using interacting particles.
problem Training latent diffusion models efficiently and accurately.
method Reformulate training as minimizing a free energy functional, then approximate with interacting particles.
result The new algorithm outperforms previous methods in experiments.
Pixel-space diffusion models outperform latent models on high-resolution image synthesis.
problem Efficiency and quality trade-off in high-resolution image synthesis.
method Sigmoid loss-weighting, simplified architecture, and resolution scaling.
result Achieved 1.5 FID on ImageNet512, new SOTA results on other datasets.
Early stopping improves sample quality in latent diffusion models.
problem Latent diffusion models degrade sample quality with conventional early stopping.
method Analyzed the interaction between latent dimension and stopping time under Gaussian framework.
result Lower-dimensional representations benefit from earlier termination, higher-dimensional spaces require later stopping.
Diffusion models reveal latent hierarchical structure in data.
problem Quantitative measurement of data's hierarchical structure.
method Forward-backward experiments in diffusion models.
result Changes in latent variables manifest as correlated chunks in data.
PaGoDA reduces diffusion model training costs by 64x.
problem Diffusion models are computationally expensive during training.
method Three-stage pipeline: downsampled training, distillation, progressive super-resolution.
result PaGoDA achieves state-of-the-art performance with reduced training costs.
New method uses Diffusion Maps for latent space modeling of dynamical systems.
problem Building reduced dynamical models from time series data.
method Two rounds of Diffusion Maps on latent coordinates, with lifting back to ambient space.
result Approximation of full state functions in reduced coordinates.
New method infers and samples point processes from latent diffusion.
problem Modeling point processes with latent diffusion.
method Itô's excursion theory for inference and sampling.
result Proposes a new method to infer and sample point processes.
Confidence-based filtering reveals latent structure in diffusion models.
problem Unclear latent structure in diffusion models.
method Confidence scores from a classifier.
result Class-relevant latent structure emerges under confidence-based filtering.
This paper proposes a method to improve VAEs by extracting latent spaces from pre-trained diffusion models.
problem VAEs struggle with generating high-quality images due to unrealistic Gaussian assumptions.
method Optimizes an encoder to maximize marginal data log-likelihood and derives a decoder analytically.
result The method enhances VAE performance by discarding Gaussian assumptions and training a separate decoder network.
VISION-XL improves HD video quality using latent image diffusion models.
problem Improving high-definition video quality and resolution.
method Latent image diffusion models and pseudo-batch consistent sampling.
result State-of-the-art video reconstruction across various inverse problems.
Latent Block-Diffusion Temporal Point Processes (LBDTPP) is a semi-autoregressive framework for generating asynchronous event sequences.
problem Generating asynchronous event sequences
method Latent Block-Diffusion Temporal Point Processes
result Outperforms state-of-the-art TPP baselines in both unconditional and conditional generation tasks
New method solves linear inverse problems using diffusion models.
problem Linear inverse problems in various domains.
method Posterior sampling with latent diffusion models.
result Provable sample recovery in linear models, outperforming previous methods.
LSSDM improves imputation of multivariate time series data.
problem Imputation of multivariate time series data without labels.
method LSSDM projects observed data into latent space, reconstructs missing values without labels, and uses a conditional diffusion model for precise imputation.
result LSSDM achieves superior imputation performance and uncertainty analysis.
CSDM integrates compressed sensing into diffusion models for faster data generation.
problem Efficiently generating synthetic data in high-dimensional spaces.
method Integrating compressed sensing into diffusion models (CSDM) to reduce dimensionality and accelerate inference.
result Achieves provably faster convergence and better latent space dimension selection.
Diffusion Transformer captures spatial-temporal dependencies in sequential data.
problem Capturing rich spatial and temporal dependencies in sequential data.
method Established theoretical guarantees for diffusion transformers learning Gaussian process data.
result Spatial-temporal dependencies are captured within attention layers of diffusion transformers.
Paper introduces Latent-CLIP for efficient text-image comparison in latent space.
problem Efficiently compare text and images in latent space without costly decoding.
method Trains CLIP model in latent space, uses Latent-CLIP rewards for noise optimization, and guides generation away from harmful content.
result Latent-CLIP matches CLIP performance on text-image classification and harmful content detection.
Improved deep hierarchical VAE with diffusion-based VampPrior.
problem Latent variable generative modeling challenges.
method Hierarchical VAE with amortized diffusion-based VampPrior.
result Better performance with fewer parameters and improved stability.
We introduce and study a class of probabilistic generative models, where the latent object is a finite-dimensional diffusion process on a finite time interval and the observed variable is drawn conditionally on the terminal point of the diffusion. We make the following contributions: We provide a unified viewpoint on b…
KIPLMC methods improve statistical inference in latent variable models.
problem Statistical inference in latent variable models.
method Joint diffusion process in parameter and latent variable spaces, with two explicit discretizations.
result KIPLMC methods achieve accelerated convergence rates in Wasserstein-2 distance.
NDMs enable non-linear transformations in diffusion models for better generative tasks.
problem Limited to linear transformations, diffusion models struggle with generative tasks.
method Presented NDMs that allow time-dependent non-linear transformations.
result NDMs outperform conventional diffusion models in likelihood and sample quality.