Deep Actor-Critics with Tight Risk Certificates
Preprint on arXiv, 2025
Recursive PAC-Bayesian bounds for actor-critic methods.
Manuel Haussmann
I'm a postdoctoral researcher at the University of Southern Denmark in Odense, where I work in the group of Melih Kandemir. I did my PhD at Heidelberg University with Fred Hamprecht and Melih, and spent two great years at Aalto University in Finland working with Samuel Kaski and Harri Lähdesmäki. My research focus is on Bayesian probabilistic machine learning with applications ranging from physics over health data to reinforcement learning. I am an action editor for TMLR, a reviewer for most major machine learning conferences (AAAI, AISTATS, ICLR, ICML, NeurIPS), and a member of the ICBINB initiative.
I'm on the job market starting from June 2026.
I'm always interested in new collaborations. Feel free to send me an email!
Research
Latent mixed-effect models for high-dimensional longitudinal data
Transactions on Machine Learning Research (TMLR); published version, 2025
Machine Learning for Health Symposium; extended version, 2024
ICML Workshop on Structured Probabilistic Inference & Generative Modeling, 2024
A structured variational auto-encoder constrained with a mixed model for longitudinal data.
Improving Plasticity in Non-stationary Reinforcement Learning with Evidential Proximal Policy Optimization
Preprint on arXiv, 2025
Fixing plasticity issues in PPO via evidential learning.
High-Dimensional Bayesian Optimisation with Gaussian Process Prior Variational Autoencoders
International Conference on Learning Representations (ICLR), 2025
Improving Bayesian optimization by incorporating prior knowledge into the latent space.
Accurate Surrogate Amplitudes with Calibrated Uncertainties
Preprint on arXiv, 2024
An evaluation of uncertainty calibration with BNNs and ensembles on LHC physics data.
Deep Exploration with PAC-Bayes
Preprint on arXiv, 2024
We propose a novel actor-critic algorithm for continuous control in sparse reward settings, using a critic ensemble and PAC-Bayesian bounds on Bellman operator error to enable deep exploration via posterior sampling.
Deterministic Uncertainty Propagation for Improved Model-Based Offline Reinforcement Learning
Conference on Neural Information Processing Systems (NeurIPS), 2024
We introduce a moment-matching based uncertainty penalty for the Bellman operator in an offline RL setting.
PAC-Bayesian Soft Actor-Critic Learning
Advances in Approximate Bayesian Inference (AABI), 2024
We introduce the first PAC Bayesian bound trainable in a modern deep reinforcement learning pipeline to solve continuous control tasks.
Latent variable model for high-dimensional point process with structured missingness
International Conference on Machine Learning (ICML), 2024
A model for longitudinal data that can handle irregular observations and model missing not at random data.
Estimating treatment effects from single-arm trials via latent-variable modeling
International Conference on Artificial Intelligence and Statistics (AISTATS), 2024
We introduce a probabilistic model to combine different patient populations for trial effect estimation.
A comparative study of clinical trial and real-world data in patients with diabetic kidney disease
Nature Scientific Reports, 2024
An evaluation on the differences between a curated medical trial data set and uncurated electronic health records data.
Practical Equivariances via Relational Conditional Neural Processes
Conference on Neural Information Processing Systems (NeurIPS), 2023
We introduce a method to incorporate high-dimensional equivariances into (conditional) neural processes.
Understanding Event-Generation Networks via Uncertainties
SciPost Physics, 2022
An evaluation of Bayesian normalizing flow for LHC events.
Evidential Turing Processes
International Conference on Learning Representations (ICLR), 2022
Evidential deep learning combined with concepts from neural Turing machines and neural processes.
Learning Partially Known Stochastic Dynamics with Empirical PAC Bayes
International Conference on Artificial Intelligence and Statistics (AISTATS), 2021
We introduce a probabilistic method to learn dynamical systems that allows for partial input incorporation.
Bayesian Evidential Deep Learning with PAC Regularization
Advances in Approximate Bayesian Inference (AABI), 2021
We combine evidential deep learning and PAC-based regularization with moment-matching approaches from Bayesian deep learning.
Deep-learning jets with uncertainties and more
SciPost Physics, 2020
We discuss Bayesian deep-learning methods for jet classification in particle physics.
Sampling-Free Variational Inference of Bayesian Neural Networks by Variance Backpropagation
Uncertainty in Artificial Intelligence Conference (UAI), 2020
A sampling-free approach to BNN learning via ReLU decomposition.
Deep Active Learning with Adaptive Acquisition
International Joint Conference on Artificial Intelligence (IJCAI), 2019
We introduce a way to adaptively learn the acquisition function for active learning with reinforcement feedback.
LeMoNADe: Learned Motif and Neuronal Assembly Detection in calcium imaging videos
International Conference on Learning Representations (ICLR), 2019
We present LeMoNADe, an end-to-end learned motif detection method directly operating on calcium imaging videos.
Variational Bayesian Multiple Instance Learning with Gaussian Processes
Conference on Computer Vision and Pattern Recognition (CVPR), 2017
Instance level inference with closed-form updates in a Gaussian processed-based model for weakly supervised multiple instance learning and object detection.
Variational Weakly Supervised Gaussian Processes
British Machine Vision Conference (BMVC), 2016
We introduce a model for multiple instance learning with Gaussian processes for bag-level predictions with closed-form updates.