Laplace Transform Based Low-Complexity Learning of Continuous Markov Semigroups

Authors: Vladimir R Kostic, Karim Lounici, Hélène Halconruy, Timothée Devergne, Pietro Novelli, Massimiliano Pontil

ICML 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, we demonstrate our theoretical findings in several experiments. Our experiments show a striking performance improvement over TO-based and other IG methods, as predicted by our theory.
Researcher Affiliation Academia 1CSML, Istituto Italiano di Tecnologia, Genova, Italy 2Faculty of Science, University of Novi Sad, Serbia 3CMAP, Ecole Polytechnique, Paris, France 4SAMOVAR, T el ecom Sud-Paris, Institut Polytechnique de Paris, France 5MODAL X, Universit e Paris Nanterre, France 6AI Center, University College London, UK.
Pseudocode Yes Algorithm 1 Primal La RRR Algorithm 2 Dual La RRR
Open Source Code Yes The implementation of the methods, as well as all experiments are available in the Git Hub repository.
Open Datasets Yes We use two independent trajectories from (N uske et al., 2017) to train and validate the model, respectively.
Dataset Splits Yes We use two independent trajectories from (N uske et al., 2017) to train and validate the model, respectively.
Hardware Specification No No specific hardware details (like GPU/CPU models, processor types, or memory amounts) are mentioned in the paper.
Software Dependencies No The paper does not provide specific software dependencies (e.g., library or solver names with version numbers) used to replicate the experiment.
Experiment Setup Yes As features we use 1000 random Fourier features, and test if we can recover eigenvalues of the non-normal linear drift. In Fig. 1 we compare to TO RRR estimator over ten trials and for two different time-discretizations. Number of samples for t = 1e 2 is n = 1e4, while for t = 1e 3 is n = 1e5.