Solving Constrained Variational Inequalities via a First-order Interior Point-based Method

Authors: Tong Yang, Michael Jordan, Tatjana Chavdarova

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

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical analyses demonstrate clear advantages of ACVI over common first-order methods.Empirically, we document two notable advantages of ACVI over popular projection-based saddlepoint methods: (i) the ACVI iterates exhibit significantly reduced rotations, as they approach the solution from the analytic center, and (ii) while projection-based methods show extensive zigzagging when hitting a constraint, ACVI avoids this, resulting in more efficient updates 5.
Researcher Affiliation Academia Tong Yang University of California, Berkeley pptmiao@berkeley.edu Michael I. Jordan University of California, Berkeley jordan@cs.berkeley.edu Tatjana Chavdarova University of California, Berkeley tatjana.chavdarova@berkeley.edu
Pseudocode Yes Algorithm 1 ACVI pseudocode.
Open Source Code Yes Link to source code: https://github.com/Chavdarova/ACVI.
Open Datasets Yes As GANs on MNIST (Lecun & Cortes, 1998) enjoy well-established metrics, we use this setup and augment it solely with linear inequalities. Fashion-MNIST: a novel image dataset for benchmarking machine learning algorithms, 2017.
Dataset Splits No No explicit percentages, sample counts, or detailed methodology for training/validation/test splits are provided for reproducibility.
Hardware Specification Yes Hardware. We used the Colab platform (https://colab.research.google.com/) and Tesla P100 GPUs. The running times are reported in App. E.
Software Dependencies No The paper mentions 'Py Torch' and 'covopt.solvers.lp' but does not provide specific version numbers for these or any other software dependencies.
Experiment Setup Yes For all the 2D problems, we set the step size of GDA, EG and OGDA to 0.1, we use k = 5 and α = 0.5 for LA-GDA, we set β = 0.08, µ 1 = 10 5, δ = 0.5 and λ0 = 0 for ACVI; and run for 50 iterations. For ACVI, we set the number of outer loop iterations to T = 20.