Likelihood Based Generative

Likelihood-based generative models aim to learn probability distributions of data to generate new samples and estimate data likelihoods. Current research focuses on improving the accuracy and efficiency of likelihood estimation, particularly addressing issues like posterior collapse in hierarchical VAEs and likelihood misestimation in other architectures such as normalizing flows and PixelCNN++. These advancements are impacting various fields, including image generation, precipitation downscaling, and outlier detection, by enabling more accurate density estimation and improved generation quality, even with limited data. The development of novel algorithms, such as Verlet flows for exact likelihood computation and data mollification techniques for improved optimization, are driving progress in this area.

Papers

October 2, 2024

Forte : Finding Outliers with Representation Typicality Estimation
Debargha Ganguly, Warren Morningstar, Andrew Yu, Vipin Chaudhary
Generative Model Synthetic Data OOD Detection Likelihood Based Generative Synthetic Text Typicality Effect

June 26, 2024

A Likelihood-Based Generative Approach for Spatially Consistent Precipitation Downscaling
Jose González-Abad
Generative Model Adversarial Loss Climate Downscaling Likelihood Loss Likelihood Based Generative

May 5, 2024

Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models
Ezra Erives, Bowen Jing, Tommi Jaakkola
Normalizing Flow Boltzmann Distribution Flow Based Generative Likelihood Based Generative Numerical Integrator

January 20, 2024

Projected Belief Networks With Discriminative Alignment for Acoustic Event Classification: Rivaling State of the Art CNNs
Paul M. Baggenstoss, Kevin Wilkinghoff, Felix Govaers, Frank Kurth
Internal State Generative Network Event Classification Likelihood Based Generative

December 30, 2023

A Bayesian Unification of Self-Supervised Clustering and Energy-Based Models
Emanuele Sansone, Robin Manhaeve
Self Supervised Learning Energy Based Model Self Supervised Objective Likelihood Based Generative Self Supervised Clustering Finite Unification Set

June 2, 2023

LIC-GAN: Language Information Conditioned Graph Generative GAN Model
Robert Lo, Arnhav Datar, Abishek Sridhar
Generative Adversarial Network Likelihood Based Generative

May 30, 2023

One-Line-of-Code Data Mollification Improves Optimization of Likelihood-based Generative Models
Ba-Hien Tran, Giulio Franzese, Pietro Michiardi, Maurizio Filippone
Generative Model Optimization Purpose Score Based Diffusion Model Likelihood Based Generative

May 22, 2023

Phased Data Augmentation for Training a Likelihood-Based Generative Model with Limited Data
Yuta Mimura
Generative Model Data Augmentation GAN Model Limited Data Data Augmentation Technique GAN Architecture Likelihood Based Generative

May 6, 2023

Improved Techniques for Maximum Likelihood Estimation for Diffusion ODEs
Kaiwen Zheng, Cheng Lu, Jianfei Chen, Jun Zhu
Diffusion Model Maximum Likelihood Probability Flow ODE Improved Technique Likelihood Based Generative Diffusion ODE

April 22, 2023

Learning Symbolic Representations Through Joint GEnerative and DIscriminative Training
Emanuele Sansone, Robin Manhaeve
Generative Question Symbolic Representation Likelihood Based Generative Joint Generative Discriminative Training

February 20, 2023

Discouraging posterior collapse in hierarchical Variational Autoencoders using context
Anna Kuzina, Jakub M. Tomczak
Context Information Hierarchical Variational Posterior Collapse Likelihood Based Generative Hierarchical VAE

December 1, 2022

GrannGAN: Graph annotation generative adversarial networks
Yoann Boget, Magda Gregorova, Alexandros Kalousis
Generative Adversarial Network Graph Representation Likelihood Based Generative Annotated Graph Implicit Distribution

September 22, 2022

Turning Normalizing Flows into Monge Maps with Geodesic Gaussian Preserving Flows
Guillaume Morel, Lucas Drumetz, Simon Benaïchouche, Nicolas Courty, François Rousseau
Optimal Transport Normalizing Flow Likelihood Based Generative Monge Map

August 29, 2022

Shaken, and Stirred: Long-Range Dependencies Enable Robust Outlier Detection with PixelCNN++
Barath Mohan Umapathi, Kushal Chauhan, Pradeep Shenoy, Devarajan Sridharan
Long Range Dependency Likelihood Based Generative Single Pixel Network

June 22, 2022

Neural Inverse Transform Sampler
Henry Li, Yuval Kluger
Density Estimation Conditional Density Likelihood Based Generative

June 7, 2022

Joint Manifold Learning and Density Estimation Using Normalizing Flows
Seyedeh Fatemeh Razavi, Mohammad Mahdi Mehmanchi, Reshad Hosseini, Mostafa Tavassolipour
Normalizing Flow Density Estimation Low Dimensional Manifold Likelihood Based Generative

Likelihood Based Generative

Papers

Forte : Finding Outliers with Representation Typicality Estimation

A Likelihood-Based Generative Approach for Spatially Consistent Precipitation Downscaling

Verlet Flows: Exact-Likelihood Integrators for Flow-Based Generative Models

Projected Belief Networks With Discriminative Alignment for Acoustic Event Classification: Rivaling State of the Art CNNs

A Bayesian Unification of Self-Supervised Clustering and Energy-Based Models

LIC-GAN: Language Information Conditioned Graph Generative GAN Model

One-Line-of-Code Data Mollification Improves Optimization of Likelihood-based Generative Models

Phased Data Augmentation for Training a Likelihood-Based Generative Model with Limited Data

Improved Techniques for Maximum Likelihood Estimation for Diffusion ODEs

Learning Symbolic Representations Through Joint GEnerative and DIscriminative Training

Discouraging posterior collapse in hierarchical Variational Autoencoders using context

GrannGAN: Graph annotation generative adversarial networks

Turning Normalizing Flows into Monge Maps with Geodesic Gaussian Preserving Flows

Shaken, and Stirred: Long-Range Dependencies Enable Robust Outlier Detection with PixelCNN++

Neural Inverse Transform Sampler

Joint Manifold Learning and Density Estimation Using Normalizing Flows