DeepSaDe: Learning Neural Networks That Guarantee Domain Constraint Satisfaction

Authors: Kshitij Goyal, Sebastijan Dumancic, Hendrik Blockeel

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Evaluation on various machine learning tasks demonstrates that our approach is flexible enough to enforce a wide variety of domain constraints and is able to guarantee them in neural networks. and Afterward, we compare our approach to related work and evaluate it experimentally. and 6 Experiments We evaluate multiple use cases in various ML tasks with complex domain constraints.
Researcher Affiliation Academia 1KU Leuven, Belgium 2Delft University of Technology, The Netherlands
Pseudocode Yes Algorithm 1: Deep Satisfiability Descent (Deep Sa De)
Open Source Code No No explicit statement about providing open-source code or a link to a code repository for the methodology described in this paper was found.
Open Datasets Yes UC4: A multi-label classification problem of identifying the labels from a sequence of 4 MNIST images.
Dataset Splits Yes and the data is split 70/20/10 into train/test/validation.
Hardware Specification Yes We ran experiments on an Intel(R) Xeon(R) Silver 4214 CPU @ 2.20GHz machine with 125 GB RAM.
Software Dependencies No The paper mentions 'Z3 solver' but does not specify its version number. No other software dependencies are listed with specific versions.
Experiment Setup No Refer to appendix A.2 for details on the architectures and hyper-parameters. The main text itself does not provide concrete hyperparameter values or detailed training configurations.