Safe Semi-Supervised Learning of Sum-Product Networks

Martin Trapp, Tamas Madl, Robert Peharz, Franz Pernkopf, Robert Trappl
Proceedings of the 33rd Conference on Uncertainty in Artificial Intelligence, PMLR R15:681-690, 2017.

Abstract

In several domains obtaining class annotations is expensive while at the same time unlabelled data are abundant. While most semi-supervised approaches enforce restrictive assumptions on the data distribution, recent work has man- aged to learn semi-supervised models in a non- restrictive regime. However, so far such ap- proaches have only been proposed for linear models. In this work, we introduce semi- supervised parameter learning for Sum-Product Networks (SPNs). SPNs are deep probabilis- tic models admitting inference in linear time in number of network edges. Our approach has several advantages, as it (1) allows genera- tive and discriminative semi-supervised learn- ing, (2) guarantees that adding unlabelled data can increase, but not degrade, the performance (safe), and (3) is computationally efficient and does not enforce restrictive assumptions on the data distribution. We show on a variety of data sets that safe semi-supervised learning with SPNs is competitive compared to state-of-the- art and can lead to a better generative and dis- criminative objective value than a purely super- vised approach.

Cite this Paper

BibTeX
@InProceedings{pmlr-vR15-trapp17a, title = {Safe Semi-Supervised Learning of Sum-Product Networks}, author = {Trapp, Martin and Madl, Tamas and Peharz, Robert and Pernkopf, Franz and Trappl, Robert}, booktitle = {Proceedings of the 33rd Conference on Uncertainty in Artificial Intelligence}, pages = {681--690}, 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/trapp17a/trapp17a.pdf}, url = {https://proceedings.mlr.press/r15/trapp17a.html}, abstract = {In several domains obtaining class annotations is expensive while at the same time unlabelled data are abundant. While most semi-supervised approaches enforce restrictive assumptions on the data distribution, recent work has man- aged to learn semi-supervised models in a non- restrictive regime. However, so far such ap- proaches have only been proposed for linear models. In this work, we introduce semi- supervised parameter learning for Sum-Product Networks (SPNs). SPNs are deep probabilis- tic models admitting inference in linear time in number of network edges. Our approach has several advantages, as it (1) allows genera- tive and discriminative semi-supervised learn- ing, (2) guarantees that adding unlabelled data can increase, but not degrade, the performance (safe), and (3) is computationally efficient and does not enforce restrictive assumptions on the data distribution. We show on a variety of data sets that safe semi-supervised learning with SPNs is competitive compared to state-of-the- art and can lead to a better generative and dis- criminative objective value than a purely super- vised approach.}, note = {Reissued by PMLR on 04 October 2026.} }
Endnote
%0 Conference Paper %T Safe Semi-Supervised Learning of Sum-Product Networks %A Martin Trapp %A Tamas Madl %A Robert Peharz %A Franz Pernkopf %A Robert Trappl %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-trapp17a %I PMLR %P 681--690 %U https://proceedings.mlr.press/r15/trapp17a.html %V R15 %X In several domains obtaining class annotations is expensive while at the same time unlabelled data are abundant. While most semi-supervised approaches enforce restrictive assumptions on the data distribution, recent work has man- aged to learn semi-supervised models in a non- restrictive regime. However, so far such ap- proaches have only been proposed for linear models. In this work, we introduce semi- supervised parameter learning for Sum-Product Networks (SPNs). SPNs are deep probabilis- tic models admitting inference in linear time in number of network edges. Our approach has several advantages, as it (1) allows genera- tive and discriminative semi-supervised learn- ing, (2) guarantees that adding unlabelled data can increase, but not degrade, the performance (safe), and (3) is computationally efficient and does not enforce restrictive assumptions on the data distribution. We show on a variety of data sets that safe semi-supervised learning with SPNs is competitive compared to state-of-the- art and can lead to a better generative and dis- criminative objective value than a purely super- vised approach. %Z Reissued by PMLR on 04 October 2026.
APA
Trapp, M., Madl, T., Peharz, R., Pernkopf, F. & Trappl, R.. (2017). Safe Semi-Supervised Learning of Sum-Product Networks. Proceedings of the 33rd Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research R15:681-690 Available from https://proceedings.mlr.press/r15/trapp17a.html. Reissued by PMLR on 04 October 2026.

Related Material