Deep Bayesian Optimization on Attributed Graphs
Authors: Jiaxu Cui, Bo Yang, Xia Hu1377-1384
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Intensive experiments are conducted on both artificial and real-world problems, including molecular discovery and urban road network design, and demonstrate the effectiveness of the DGBO compared with the state-of-the-art. |
| Researcher Affiliation | Academia | 1College of Computer Science and Technology, Jilin University, Changchun, China 2Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education, China 3Department of Computer Science and Engineering, Texas A&M University, College Station, United States |
| Pseudocode | Yes | Algorithm 1: DGBO |
| Open Source Code | No | The paper does not provide any explicit statement about releasing source code for the DGBO methodology described, nor does it include a link to a code repository. |
| Open Datasets | Yes | Delaney is a molecular data set having 1,122 molecules whose aqueous solubility has been measured by (Delaney 2004). ZINC includes 20,000 drug-like commercially available molecules extracted at random from the ZINC database (Irwin et al. 2012). Sioux Falls. This data set is widely used in transportation studies (Leblanc, Morlok, and Pierskalla 1975). |
| Dataset Splits | No | The paper describes an iterative Bayesian optimization process where the surrogate model is retrained on an ever-growing set of evaluated graphs (e.g., 'D0= {(G1, y1), (G2, y2), ..., (GM, y M)}' and 'augment data Dt = Dt 1 {(Gnext, ynext)}'). It does not specify fixed, pre-defined train/validation/test dataset splits for the overall problem or the main datasets used in the experiments. |
| Hardware Specification | No | The paper mentions that 'all methods on the same hardware setting' were used for comparison, and refers to 'massive computing resources' for evaluations, but it does not provide any specific details about the hardware specifications (e.g., CPU/GPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper mentions software and algorithms used, such as 'Network X tool' for generating synthetic data, 'Adam' for optimization, 'MCMC sampler' and 'Frank Wolfe algorithm', but it does not specify any version numbers for these software dependencies. |
| Experiment Setup | Yes | Table 1 lists specific optimal surrogate architecture parameters found via Bayesian optimization, including '# GC layers', '# FC layers', '# units of GC', '# units of pooling', '# units of FC', 'σ(.) of GC', 'σ(.) of pooling', 'σ(.) of FC', 'Learning rate', 'Dropout', and 'Penalty coefficient'. Additionally, it states, 'both the number of initializing graphs M and the iterations of retraining Re are set to 20, B is set to 4, and all algorithms run 5 times to eliminate random effects.' |