Online Learning under Adversarial Nonlinear Constraints

Authors: Pavel Kolev, Georg Martius, Michael Muehlebach

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We also empirically evaluate our algorithm on two-player games, where the players are subjected to a shared constraint. We apply our algorithm and show numerical experiments that support our theoretical findings.
Researcher Affiliation Academia 1Max Planck Institute for Intelligent Systems, Tübingen, Germany 2University of Tübingen, Tübingen, Germany
Pseudocode Yes Algorithm 1 Constraint Violation Velocity Projection (CVV-Pro)
Open Source Code No The paper states 'For the implementation of CVV-Pro we have used the MATLAB R2019a numerical computing software.' but does not provide a link or explicit statement about the release of their source code.
Open Datasets No The paper describes generating random instances for the two-player game simulation, stating 'Each component of the utility matrix A Rn n is sampled from the normal distribution and the constraint matrices Cx, Cy [0, 1]m n have each of their components sampled uniformly at random from [0, 1]'. No publicly available dataset is used or linked.
Dataset Splits No The paper describes numerical simulations with generated data but does not specify training, validation, or test dataset splits.
Hardware Specification Yes The computation of the experiment takes about 4 hours on a machine with CPU: Intel(R) i7-6800K 3.40 GHz with 6 cores, GPU: NVIDIA Ge Force GTX 1080, and RAM: 32 GB.
Software Dependencies Yes For the implementation of CVV-Pro we have used the MATLAB R2019a numerical computing software.
Experiment Setup Yes We implemented our algorithm with ηt = 1/(αt) and α = 100. We report results from numerical simulations with decision dimension n = 100, m = 10 shared resource constraints, T = 4000 iterations, and five independently sampled instances of the two-player game.