Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1]
Using Collective Behavior of Coupled Oscillators for Solving DCOP
Authors: Allan R. Leite, Fabricio Enembreck
JAIR 2019 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In order to evaluate the performance of the proposed algorithm on large instances of DCOP, our experiments considered two types of problems: random problems and realistic meeting scheduling problems. Experiments were performed in FRODO framework on a machine with an Intel(R) Core(TM) i7-7500U CPU with 2.9 Ghz and 16GB of RAM. ... Our experimental evaluation benchmarks COOPT with state-of-the-art incomplete DCOP algorithms, named DSA (Fitzpatrick & Meertens, 2003), DSA-SDP (Zivan et al., 2014), DUCT (Ottens et al., 2017), GDBA (Okamoto et al., 2016), MGM (Zhang et al., 2005), and MGM-2 (Pearce et al., 2008). ... The results were averaged over 30 executions for each algorithm and problem instance. We perform a performance evaluation using well-known computation and communication metrics, such as number of messages, amount of exchanged information (Modi et al., 2005), and non-concurrent constraint checks (NCCC) (Meisels, Kaplansky, Razgon, & Zivan, 2002). Because all evaluated algorithms are incomplete methods, our performance analysis also compares solution costs. |
| Researcher Affiliation | Academia | Allan R. Leite EMAIL Department of Applied Informatics Pontifical Catholic University of Parana Imaculada Concei c ao St., 1155, Curitiba-PR, Brazil. Fabr ıcio Enembreck EMAIL Department of Applied Informatics Pontifical Catholic University of Parana Imaculada Concei c ao St., 1155, Curitiba-PR, Brazil |
| Pseudocode | Yes | Algorithm 1 Procedures of COOPT |
| Open Source Code | No | The paper does not provide concrete access to source code for the methodology described. It mentions using the FRODO framework (a third-party tool) but does not state that the COOPT algorithm's implementation is publicly available. |
| Open Datasets | No | For random problems, the paper states: 'We generated 30 instances of random problems for each combination of variables, densities, and topologies.' For meeting scheduling problems, it states: 'We generated 30 instances following these setups, where each DCOP was composed by 220 agents and 400 variables with a single domain size |D| = 20.' The paper describes how problem instances were generated but does not provide concrete access information (link, DOI, repository, or formal citation for a public dataset) for these instances. |
| Dataset Splits | No | The paper describes generating 30 instances of random problems and 30 instances of meeting scheduling problems, and then evaluating algorithms by averaging over 30 executions for each algorithm and problem instance. However, it does not specify any training, testing, or validation splits for these datasets. |
| Hardware Specification | Yes | Experiments were performed in FRODO framework on a machine with an Intel(R) Core(TM) i7-7500U CPU with 2.9 Ghz and 16GB of RAM. |
| Software Dependencies | No | The paper mentions that 'Experiments were performed in FRODO framework' and Figure 1 notes 'Generated from Mathematica Software (Wolfram, 2003)'. However, it does not provide specific version numbers for FRODO or Mathematica, nor does it list other key software components with their versions. |
| Experiment Setup | Yes | Our experimental evaluation benchmarks COOPT with state-of-the-art incomplete DCOP algorithms... We also assume n = 100 iterations for all local search algorithms. ... DSA uses an activation probability p before choosing new assignments, and we considered p = 0.7... DSA-SDP has four parameters p A, p B, p C, and p D which represent the potential of improvement for calculating the probability of choosing new assignments, so we use values p A = 0.6, p B = 0.15, p C = 0.4, and p D = 0.8... DUCT, its termination condition occurs when the difference of the best value and the minimal lower bound is smaller than the ϵ confidence bound, so we assumed ϵ = 0.05... GDBA with multiplicative weight manner (M), non-minimum constraint violation (NM), and table scope of cost increase (T), resulting in (M, NM, T) combination... COOPT requires a parameter K that represents the global coupling strength among connected variables... we assumed a coupling strength K = 1 in our evaluation... |