Yingzhen Li

Associate Professor@Imperial College London·英国·OpenReview

研究方向

causal representation learning · continue learning · disentangled representation · Gaussian process · meta learning · sequential generative models · stochastic gradient MCMC · adversarial attacks and defences · score matching · Stein's method · Bayesian neural networks · deep generative models · transfer learning · approximate inference · variational inference · message passing

Yingzhen Li

Inference-Time Scaling of Discrete Diffusion Models via Importance Weighting and Optimal Proposal Design

Unsupervised Parallel MRI Reconstruction via Projected Conditional Flow Matching

Memory-Driven Self-Improvement with Large Language Models

Neural Flow Samplers with Shortcut Models

HiBBO: HiPPO-based Space Consistency for High-dimensional Bayesian Optimisation

Improving Probabilistic Diffusion Models With Optimal Diagonal Covariance Matching

Compact Memory for Continual Logistic Regression

Neural Stochastic Flows: Solver-Free Modelling and Inference for SDE Solutions

Recurrent Memory for Online Interdomain Gaussian Processes

Discrete Neural Flow Samplers with Locally Equivariant Transformer

Variational Uncertainty Decomposition for In-Context Learning

Causal Discovery from Conditionally Stationary Time Series

Neural Mutual Information Estimation with Reference Distributions

TabUnite: Efficient Encoding Schemes for Flow and Diffusion Tabular Generative Models

From Patches to Graphs: Towards Image Diffusion Models with GNNs