Fuzzy Inference Systems (FIS) effectively model and reason with complex and uncertain information in an interpretable, understandable, and transparent way. Takagi-Sugeno-Kang (TSK) is one of the most widespread types of FIS, appreciated for its ability to output crisp values by leveraging linear models as consequents. In this tribute to Prof. Sugeno, we discuss our previous works based on TSK inference. In particular, we focus on pyFUME, a Python library for automatically estimating first-order TSK FIS from data. We introduce a relevant advancement in pyFUME, i.e., the ability to handle categorical variables within the consequents, which significantly enhances the model’s performance in regression and classification tasks. This improved version completes the foundation for pyFUME to handle mixed-type data. Our results on three diverse datasets show the capabilities of our method, which performs better than the previous implementation.
Bacciu, L., Cazzaniga, P., Gallese, C., Fuchs, C., Kaymak, U., Papetti, D., et al. (2025). Our Fruitful Relationship with Sugeno Inference, from FUMOSO to pyFUME. In Information Processing and Management of Uncertainty in Knowledge-Based Systems 20th International Conference, IPMU 2024, Lisbon, Portugal, July 22-26, 2024, Proceedings, Volume 2 (pp.12-24). Springer Science and Business Media Deutschland GmbH [10.1007/978-3-031-74000-8_2].
Our Fruitful Relationship with Sugeno Inference, from FUMOSO to pyFUME
Papetti D. M.;
2025
Abstract
Fuzzy Inference Systems (FIS) effectively model and reason with complex and uncertain information in an interpretable, understandable, and transparent way. Takagi-Sugeno-Kang (TSK) is one of the most widespread types of FIS, appreciated for its ability to output crisp values by leveraging linear models as consequents. In this tribute to Prof. Sugeno, we discuss our previous works based on TSK inference. In particular, we focus on pyFUME, a Python library for automatically estimating first-order TSK FIS from data. We introduce a relevant advancement in pyFUME, i.e., the ability to handle categorical variables within the consequents, which significantly enhances the model’s performance in regression and classification tasks. This improved version completes the foundation for pyFUME to handle mixed-type data. Our results on three diverse datasets show the capabilities of our method, which performs better than the previous implementation.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
