Risk-Averse No-Regret Learning in Online Convex Games

Authors: Zifan Wang, Yi Shen, Michael Zavlanos

ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We illustrate our method on an online market problem that we model as a Cournot game (Allaz & Vila, 1993). ... In section 5, we use an online market example to illustrate the effectiveness of the proposed algorithms. ... 5. Numerical Experiments
Researcher Affiliation Academia 1School of Electrical Engineering and Computer Science, KTH Royal Institute of Technology, Stockholm, Sweden. 2Department of Mechanical Engineering & Material Science, Duke University, Durham, NC 27708, USA.
Pseudocode Yes Algorithm 1 Risk-averse learning ... Algorithm 2 Risk-averse learning with sample reuse ... Algorithm 3 Risk-averse learning with residual feedback
Open Source Code No The paper does not provide any links or explicit statements about the availability of open-source code for the described methodology.
Open Datasets No The paper describes a synthetic Cournot game setup with a defined cost function and a uniform random variable for uncertainty, rather than using a publicly available or open dataset. For example: "The cost term ξixi models the uncertainty in the market, which is proportional to production."
Dataset Splits No The paper describes a numerical simulation setup for a Cournot game and does not specify training, validation, or test dataset splits, as it's not based on pre-existing data splits.
Hardware Specification No The paper does not explicitly describe the hardware used to run its experiments.
Software Dependencies No The paper does not provide specific version numbers for any software dependencies used in the experiments.
Experiment Setup Yes In Algorithm 1, 2, and 3, the requirements are listed as: "Initial value x0, step size η, parameters a, b, δ, T, risk level αi, i = 1, , N.". In Section 5, "Numerical Experiments", it states: "We let α0 = 0.5 and α1 = 0.3, i.e, firm 1 is more risk sensitive than firm 0." and "We implement a hybrid sampling strategy for Algorithm 2 and select the switching time step as 15000". It also mentions: "All other parameters in Algorithms 1, 2 and 3 are tuned so that the three algorithms achieve individually their best performance."