Lu Yuan, currently living in London.
about
Hello! I am a machine learning researcher currently living in London. I work on diffusion models, representation learning, ebms and reinforcement learning. I also worked on Gaussian process and theoretical neuroscience in the past. Currently, my focus is deep learning, neural networks and core ML theory.
right now
- working with Prof. Dongsheng Li, Dr. Dongqi Han and Dr. Yansen Wang on diffusion planners and energy-based models at Microsoft Research
- working with Prof. Yue Song, on Helmholtz flows matching at Caltech/Tsinghua
education
- MSc Machine Learning, UCL (Distinction) - 2025
- BSc Psychology, UCL (1st class) - 2023
before this
- BA Cinematography and Film Production, Beijing Film Academy (Dropout) - 2019
worked for/with
- Prof. David Barber - MSc Thesis, One-Step Generation via Diffusive Auxiliary Variational Autoencoders, UCL - 2025
- Prof. Jinwen Ma - Research Assistant, EM algorithm and Sparse Gaussian Process, Peking University - 2024
- Prof. Maarten Speekenbrink - BSc Thesis, Reinforcement Learning under Uncertainty in Game Theory, UCL - 2023
- Prof. Fred Dick - Research Assistant, Auditory Neuroscience, UCL - 2022
- Prof. Fred Dick and Dr Magdalena Kachlicka, Research Project, Representation Learning for Natural Sound Categories, UCL - 2022
exhibitions
- Film Project - Beijing - Three Shadows Photography Art Centre - 2020
my face
publishings
projects
-
One Step Diffusion via Auxiliary Variational Autoencoders Unify diffusion modelling and variational autoencoders into a single-step generative framework
2025deep generative models, diffusion models, Langevin dynamics
-
Hamiltonian Variational Auto-Encoder Combining Hamiltonian Importance Sampling with Variational Auto-Encoders.
2024deep generative models, importance sampling, hamiltonian monte carlo, normalizing flow
-
Modelling Inequity Aversion with Multi-Agent Reinforcement Learning Exploring population dynamics of agents with inequity aversion in sequential social dilemmas.
2024reinforcement learning, game theory, social psychology
-
Dope of Dopamine We used a house of distributional RL algorithms (e.g., QR, C51) to play games.
2023reinforcement learning, distributional RL, uncertainty, decision
-
Opponent Modelling in Dyadic Games Building games and inverse RL algorithms to study how humans infer uncertainty to guides their predictions of an opponent’s behavior and make actions.
2023reinforcement learning, inverse-RL, uncertainty, decision, social psychology, BSc Thesis
- >> all projects
in progress / a blog
-
Kernel Methods A walk through for kernel method in machine learning and RKHS
2024machine learning, kernels, rkhs
-
Probabilistic and Unsupervised Learning Self-study blog for the PGM course at CMU
2024graphical models, bayesian machine learning, machine learning
-
Approximate Inference Approximate inference and learning in probabilistic graphical models
2024graphical models, bayesian networks, markov networks, mcmc
-
Machine Learning Seminar Current research in machine learning.
2023machine learning, seminar
- >> all posts
notes
-
Theoretical Neuroscience Notes for Computational and System Neuroscience
2024theoretical neuroscience, computational biology
-
Convex Optimization Revision notes for convex optimization
2023optimization, machine learning
-
Reinforcement Learning RL notes and cheats sheets over the years
2022reinforcement learning, inverse RL, distributional RL
- >> all notes
photography
- presence places I lived and traveled to.
- land and sea landscapes.
- /\ "I wanted to cut up the blue of the sky."
- faces people I met, and their ideas.
- wars all over the world.
- alone together modern loneliness.
contact
- meet me in London for coffee or beer, I am in Beijing & Shanghai quite often, too.
- edluyuan /at/ gmail /dot/ com
- google scholar
- github
- linkedin
- twitter