Learning MAX-SAT from Contextual Examples for Combinatorial Optimisation
Authors: Mohit Kumar, Samuel Kolb, Stefano Teso, Luc De Raedt4493-4500
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments We answer the following research questions: Q1) Does HASSLE acquire accurate, low-regret models? Q2) Do more examples lead to a better model? Q3) How well does it scale to more complex target models? To this end, we used our MILP encoding for recovering synthetic and benchmark MAX-SAT models M of increasing complexity from examples. The experimental setup can be found at https://github.com/mohit KULeuven/hassle. |
| Researcher Affiliation | Academia | 1KU Leuven, 2University of Trento, Italy {mohit.kumar, samuel.kolb, luc.deraedt}@cs.kuleuven.be, stefano.teso@unitn.it |
| Pseudocode | No | The paper includes a 'Simplified MILP encoding' in Figure 2, which is a mathematical formulation rather than structured pseudocode or an algorithm block with explicit steps. |
| Open Source Code | Yes | The experimental setup can be found at https://github.com/mohit KULeuven/hassle. |
| Open Datasets | Yes | To make sure that HASSLE can learn hard combinatorial problems, we instead chose five phase-transition SAT instances (Gent and Walsh 1994) with 20 variables and 91 constraints 5, and for each of them, turned most of the hard constraints into soft ones by adding random weights. |
| Dataset Splits | No | The paper states 'A dataset S was collected for each model M by sampling... and then taking s+ positive and s negative examples for each context' and mentions assessing generalization across contexts. However, it does not provide specific details on training, validation, or test dataset splits (e.g., percentages, sample counts, or cross-validation setup). |
| Hardware Specification | Yes | GUROBI was used to solve the MILP encoding on an Intel(R) Xeon(R) CPU E31225 v3 @ 3.20GHz with 32 Gi B RAM. |
| Software Dependencies | No | The paper mentions the use of 'GUROBI' as a solver, but it does not specify its version number or any other software dependencies with their respective versions. |
| Experiment Setup | Yes | So to answer Q2 we did experiments by fixing some parameters (n = 10, k = 2, mhard = 5, msoft = 5, |Ψ| = 20) and varying the number of examples (s+, s ) used to learn the model, see the right graph in Figure 3. |