$\mathttVITS$ : Variational Inference Thompson Sampling for contextual bandits

Authors: Pierre Clavier, Tom Huix, Alain Oliviero Durmus

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, we demonstrate experimentally the effectiveness of VITS on both synthetic and real world datasets.
Researcher Affiliation Academia 1CMAP, CNRS, Ecole Polytechnique, Institut Polytechnique de Paris, 91120 Palaiseau, France. 2Inria Paris, 75015 Paris, France 3Centre de Recherche des Cordeliers, INSERM, Universite de Paris, Sorbonne Universite, 75006 Paris, France .
Pseudocode Yes Algorithm 1 VITS algorithm; Algorithm 2 VITS-I; Algorithm 3 VITS II / VITS II Hessian-free
Open Source Code No The paper does not provide a direct statement or link for open-source code for the described methodology.
Open Datasets Yes Finally, our last contribution is to illustrate the empirical performances of our method on a synthetic and on the real world dataset Movie Lens (Lam & Herlocker).
Dataset Splits No The paper mentions using synthetic and Movie Lens datasets, but it does not specify explicit training/validation/test splits by percentages or sample counts for reproducibility.
Hardware Specification Yes In this work, we use GPUs v100-16g or v100-32g for running our code with GPU Nvidia Tesla V100 SXM2 16 Go and CPUs with 192 Go per node.
Software Dependencies No The paper mentions re-implementing an algorithm in JAX but does not specify a version number for JAX or any other software dependencies.
Experiment Setup Yes Parameters: step-size ht, number of iterations Kt; Table 1. Lin TS hyperparameter grid-search; Table 2. LMC-TS hyperparameter grid-search; Table 3. VITS hyperparameter grid-search