Generalizing Graph Matching beyond Quadratic Assignment Model
Authors: Tianshu Yu, Junchi Yan, Yilin Wang, Wei Liu, baoxin Li
NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We deliver solvers based on two specific instances of Separable Functions, and the state-of-the-art performance of our method is verified on popular benchmarks. Three popular benchmarks are used including Random Graph Matching [10], CMU house sequence [36] and Caltech-101/MSRC object matching [10]. |
| Researcher Affiliation | Collaboration | Tianshu Yu Arizona State University tianshuy@asu.edu Junchi Yan Shanghai Jiao Tong University yanjunchi@sjtu.edu.cn Yilin Wang Arizona State University yilwang@adobe.com Wei Liu Tecent AI Lab wl2223@columbia.edu Baoxin Li Arizona State University baoxin.li@asu.edu |
| Pseudocode | Yes | Algorithm 1 Path following for GGM. Algorithm 2 Multiplication Strategy for GGM. |
| Open Source Code | No | The paper does not provide an explicit statement or link indicating the availability of open-source code for the described methodology. |
| Open Datasets | Yes | Three popular benchmarks are used including Random Graph Matching [10], CMU house sequence [36] and Caltech-101/MSRC object matching [10]. 30 pairs of images are included in this test collected from Caltech-101 [37] and MSRC3. |
| Dataset Splits | No | The paper describes how graphs are generated for evaluation and the pairing for benchmark tests, but does not specify explicit training, validation, and testing dataset splits (e.g., percentages or fixed sample counts) in a way that implies a fixed dataset partitioning for reproducibility. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used for experiments (e.g., CPU/GPU models, memory specifications). |
| Software Dependencies | No | The paper does not specify particular software dependencies with version numbers (e.g., 'Python 3.8', 'PyTorch 1.9'). |
| Experiment Setup | Yes | For path following strategy of GGM , we set θ0 = 2, α = 0.5, k = 0.2. Then if E(t) E(t 1) < η, where η is a small positive value, we identify the convergence of the iteration. If there is no such t, the algorithm stops when reaching the pre-defined maximal iteration number. In all the following experiments, we let η = 10 8. The parameter σs is empirically set to be 0.15. where a S ij measures the Euclidean distance between point i and j, and σ2 s = 2500. |