Adversarial Attacks on Linear Contextual Bandits
Authors: Evrard Garcelon, Baptiste Roziere, Laurent Meunier, Jean Tarbouriech, Olivier Teytaud, Alessandro Lazaric, Matteo Pirotta
NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We empirically validate the proposed approaches in synthetic and real-world datasets. |
| Researcher Affiliation | Industry | Evrard Garcelon Facebook AI Research evrard@fb.com Baptiste Roziere Facebook AI Research broz@fb.com Laurent Meunier Facebook AI Research laurentmeunier@fb.com Jean Tarbouriech Facebook AI Research jtarbouriech@fb.com Olivier Teytaud Facebook AI Research oteytaud@fb.com Alessandro Lazaric Facebook AI Research lazaric@fb.com Matteo Pirotta Facebook AI Research pirotta@fb.com |
| Pseudocode | Yes | Figure 1: Contextual ACE algorithm. Figure 2: Conic Attack algorithm. |
| Open Source Code | No | No explicit statement about releasing code for the described methodology or a link to a source code repository was found. |
| Open Datasets | Yes | In this section, we conduct experiments on the attacks on contextual bandit problems with simulated data and two real-word datasets: Movie Lens25M [32] and Jester [33]. |
| Dataset Splits | No | The paper does not explicitly provide specific details about train/validation/test dataset splits, such as percentages, sample counts, or explicit splitting methodology. It mentions using synthetic data and real-world datasets (Movie Lens25M, Jester) and running simulations. |
| Hardware Specification | No | No specific hardware details (e.g., CPU/GPU models, memory, or cluster specifications) used for running the experiments were provided in the paper. |
| Software Dependencies | No | The paper does not explicitly list specific software dependencies with their version numbers (e.g., Python 3.x, PyTorch 1.y) that would allow for replication of the experimental environment. |
| Experiment Setup | Yes | As parameters, we use L = 1 for the maximal norm of the contexts, δ = 0.01, υ = σ (1 + L2(1 + |Sta|)/λ)/δ , |