Lifted Inference for Convex Quadratic Programs
Authors: Martin Mladenov, Leonard Kleinhans, Kristian Kersting
AAAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our intention here is to investigate whether (Q) AI can potentially benefit from relational and lifted QPs? As our main experiment, we compared our lifted QP approach to relational linear programs (Kersting, Mladenov, and Tokmakov 2015), following their experimental setup for the Cora dataset (Sen et al. 2008) consisting of 2708 scientific papers classified into seven classes. |
| Researcher Affiliation | Academia | Martin Mladenov TU Dortmund University, Germany martin.mladenov@cs.tu-dortmund.de Leonard Kleinhans TU Dortmund University, Germany leonard.kleinhans@cs.tu-dortmund.de Kristian Kersting TU Dortmund University, Germany kristian.kersting@cs.tu-dortmund.de |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. Figure 1 (left) shows a mathematical formulation of a QP, not an algorithm. |
| Open Source Code | No | The paper does not provide any concrete access to source code, such as a repository link or an explicit statement about code release for the methodology described. |
| Open Datasets | Yes | As our main experiment, we compared our lifted QP approach to relational linear programs (Kersting, Mladenov, and Tokmakov 2015), following their experimental setup for the Cora dataset (Sen et al. 2008) consisting of 2708 scientific papers classified into seven classes. |
| Dataset Splits | Yes | We first randomly split the dataset into a labeled set L and an unlabeled test set B, according to t. Then, we split L randomly in half, leaving one half A for training, the other half becoming a validation set C. The validation set was used to select the parameters of the TC-QPSVM in a 5-fold cross-validation fashion. That is, we split the validation set into 5 subsets Ci of equal size. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, memory amounts, or detailed computer specifications) used for running its experiments. It only implies that experiments were run to measure time and performance. |
| Software Dependencies | No | The paper mentions "software packages for convex optimization such as CVXPY (Diamond, Chu, and Boyd 2014)" but does not specify any version numbers for CVXPY or other software used in the experiments. |
| Experiment Setup | Yes | The validation set was used to select the parameters of the TC-QPSVM in a 5-fold cross-validation fashion. ... For approximate lifting we used k-Means using the Euclidean metric and 500 anchor points. ... We used a grid search together with CV for selecting γ t0.25, 0.50, 1.00, 2.00, 4.00u and C t0.5, 1.0, 2.0u. |