Robust multi-label sum-product networks for binary relevance problems (multilabel-cspn)

This code was used to get experiments results for the paper "Skeptical binary inferences in multi-label problems with sets of probabilities".

This implementation uses some piece of code from the repository github:

• https://github.com/AlCorreia/SUM2019

SUM2019 implements a scalable and robust sum-product network for classification problems, based on an implementation of LearnSPN [1] in R. with a few important contributions:

1. Robust estimates [2],
2. Class-selective SPNs,
3. Memory caches.

For further details, we refer to package.

Installation / Usage

To clone the repo:

$ git clone https://github.com/salmuz/multilabel_cspn.git

To usage:

The experiments in the paper can be reproduced with the R/mlc.resampling.r script, but it is necessary to generate the datasets by using the script:

https://github.com/salmuz/classifip/blob/master/experiments/classification/mlc/mlc_resampling.py

Author

• Yonatan-Carlos Carranza-Alarcon

Example

Very soon

References

[1] Gens, Robert, and Domingos Pedro.
"Learning the structure of sum-product networks."
International conference on machine learning. 2013.

[2] Mauá, Denis Deratani, et al.
"Robustifying sum-product networks."
International Journal of Approximate Reasoning 101 (2018): 163-180.

[3] Correia, Alvaro HC, and Cassio P. de Campos. "Towards scalable and robust sum-product networks."
International Conference on Scalable Uncertainty Management. Springer, Cham, 2019.

---

If this material has been useful to you, please consider citing

Yonatan Carlos Carranza Alarcón, Sébastien Destercke. "Skeptical binary inferences in multi-label problems with sets of probabilities"

@misc{https://doi.org/10.48550/arxiv.2205.00662,
  doi = {10.48550/ARXIV.2205.00662},
  url = {https://arxiv.org/abs/2205.00662},
  author = {Alarcón, Yonatan Carlos Carranza and Destercke, Sébastien},
  keywords = {Machine Learning (stat.ML), Artificial Intelligence (cs.AI), Machine Learning (cs.LG), Combinatorics (math.CO), FOS: Computer and information sciences, FOS: Computer and information sciences, FOS: Mathematics, FOS: Mathematics},
  title = {Skeptical binary inferences in multi-label problems with sets of probabilities},
  publisher = {arXiv},
  year = {2022}
}