Adversarial Sets for Regularising Neural Link Predictors

Pasquale Minervini, Thomas Demeester, Tim Rocktäschel, Sebastian Riedel
Proceedings of the 33rd Conference on Uncertainty in Artificial Intelligence, PMLR R15:61-70, 2017.

Abstract

In adversarial training, a set of models learn together by pursuing competing goals, usually defined on single data instances. However, in relational learning and other non-i.i.d domains, goals can also be defined over sets of instances. For example, a link predictor for the IS-A relation needs to be consistent with the transitivity property: if IS-A(x1, x2) and IS-A(x2, x3) hold, IS-A(x1, x3) needs to hold as well. Here we use such assumptions for deriving an inconsistency loss, measuring the degree to which the model violates the assumptions on an adversarially-generated set of examples. The training objective is defined as a minimax problem, where an adversary finds the most offending adversarial examples by maximising the inconsistency loss, and the model is trained by jointly minimising a supervised loss and the inconsistency loss on the adversarial examples. This yields the first method that can use function-free Horn clauses (as in Datalog) to regularise any neural link predictor, with complexity independent of the domain size. We show that for several link prediction models, the optimisation problem faced by the adversary has efficient closed-form solutions. Experiments on link prediction benchmarks indicate that given suitable prior knowledge, our method can significantly improve neural link predictors on all relevant metrics.

Cite this Paper

BibTeX
@InProceedings{pmlr-vR15-minervini17a, title = {Adversarial Sets for Regularising Neural Link Predictors}, author = {Minervini, Pasquale and Demeester, Thomas and Rockt{\"a}schel, Tim and Riedel, Sebastian}, booktitle = {Proceedings of the 33rd Conference on Uncertainty in Artificial Intelligence}, pages = {61--70}, year = {2017}, editor = {Elidan, Gal and Kersting, Kristian}, volume = {R15}, series = {Proceedings of Machine Learning Research}, month = {11--15 Aug}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r15/main/assets/minervini17a/minervini17a.pdf}, url = {https://proceedings.mlr.press/r15/minervini17a.html}, abstract = {In adversarial training, a set of models learn together by pursuing competing goals, usually defined on single data instances. However, in relational learning and other non-i.i.d domains, goals can also be defined over sets of instances. For example, a link predictor for the IS-A relation needs to be consistent with the transitivity property: if IS-A(x1, x2) and IS-A(x2, x3) hold, IS-A(x1, x3) needs to hold as well. Here we use such assumptions for deriving an inconsistency loss, measuring the degree to which the model violates the assumptions on an adversarially-generated set of examples. The training objective is defined as a minimax problem, where an adversary finds the most offending adversarial examples by maximising the inconsistency loss, and the model is trained by jointly minimising a supervised loss and the inconsistency loss on the adversarial examples. This yields the first method that can use function-free Horn clauses (as in Datalog) to regularise any neural link predictor, with complexity independent of the domain size. We show that for several link prediction models, the optimisation problem faced by the adversary has efficient closed-form solutions. Experiments on link prediction benchmarks indicate that given suitable prior knowledge, our method can significantly improve neural link predictors on all relevant metrics.}, note = {Reissued by PMLR on 04 October 2026.} }
Endnote
%0 Conference Paper %T Adversarial Sets for Regularising Neural Link Predictors %A Pasquale Minervini %A Thomas Demeester %A Tim Rocktäschel %A Sebastian Riedel %B Proceedings of the 33rd Conference on Uncertainty in Artificial Intelligence %C Proceedings of Machine Learning Research %D 2017 %E Gal Elidan %E Kristian Kersting %F pmlr-vR15-minervini17a %I PMLR %P 61--70 %U https://proceedings.mlr.press/r15/minervini17a.html %V R15 %X In adversarial training, a set of models learn together by pursuing competing goals, usually defined on single data instances. However, in relational learning and other non-i.i.d domains, goals can also be defined over sets of instances. For example, a link predictor for the IS-A relation needs to be consistent with the transitivity property: if IS-A(x1, x2) and IS-A(x2, x3) hold, IS-A(x1, x3) needs to hold as well. Here we use such assumptions for deriving an inconsistency loss, measuring the degree to which the model violates the assumptions on an adversarially-generated set of examples. The training objective is defined as a minimax problem, where an adversary finds the most offending adversarial examples by maximising the inconsistency loss, and the model is trained by jointly minimising a supervised loss and the inconsistency loss on the adversarial examples. This yields the first method that can use function-free Horn clauses (as in Datalog) to regularise any neural link predictor, with complexity independent of the domain size. We show that for several link prediction models, the optimisation problem faced by the adversary has efficient closed-form solutions. Experiments on link prediction benchmarks indicate that given suitable prior knowledge, our method can significantly improve neural link predictors on all relevant metrics. %Z Reissued by PMLR on 04 October 2026.
APA
Minervini, P., Demeester, T., Rocktäschel, T. & Riedel, S.. (2017). Adversarial Sets for Regularising Neural Link Predictors. Proceedings of the 33rd Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research R15:61-70 Available from https://proceedings.mlr.press/r15/minervini17a.html. Reissued by PMLR on 04 October 2026.

Related Material