Optimization Algorithm Design via Electric Circuits
Authors: Stephen Boyd, Tetiana Parshakova, Ernest Ryu, Jaewook J. Suh
NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we use our methodology to obtain a new algorithm and experiment with it on a specific problem instance. ... We apply DADMM+C to the decentralized optimization problem and observe a speedup as shown in the right columns of Figure 9. The relative error for DADMM+C decreases to 10 10 in 66 iterations, for DADMM in 87 iterations and for P-EXTRA in 294 iterations. |
| Researcher Affiliation | Academia | Stephen P. Boyd Stanford University Tetiana Parshakova Stanford University Ernest K. Ryu UCLA Jaewook J. Suh Rice University |
| Pseudocode | No | The paper presents algorithms and equations in text and mathematical notation, but it does not include explicitly labeled pseudocode or algorithm blocks/figures. |
| Open Source Code | Yes | We provide an open-source package that implements automatic discretization of our circuits: https://github.com/cvxgrp/optimization_via_circuits |
| Open Datasets | No | We set m = 100, N = 6, S = {4, 5}, and sample vectors bi R100 from the uniform distribution over [ 100, 100]100. ... The paper does not provide access to a public dataset; it describes the generation of synthetic data for its experiments. |
| Dataset Splits | No | The paper does not explicitly mention train, validation, or test dataset splits or percentages. It uses synthetic data for its experiments. |
| Hardware Specification | No | The NeurIPS Paper Checklist states 'We use minimal CPU computation for toy experiments.' (Point 8). This is a general statement and does not specify any particular CPU models, memory, or other hardware details. |
| Software Dependencies | No | The paper mentions specific software tools like 'Ipopt [155, 9] solver' and 'PEPit [76] package', and refers to a custom package 'ciropt'. However, it does not provide specific version numbers for these software dependencies, which are necessary for full reproducibility. |
| Experiment Setup | Yes | Example. Consider the following example circuit for the minimization of a convex function f. Let R1 = R2 = R3 = 1, and C1 = C2 = 10. With our automatic discretization methodology, we find the sufficiently dissipative parameters η = 6.66, h = 6.66, α = 0, β = 1. ... We consider the circuit with R = 0.8, L = 2 and C = 15. ... The sufficiently dissipative parameters we find are η = 3.70, h = 3.52, ρ = 0, α = 0, β = 1, γ = 4.48. |