PyTSK: A Python Toolbox for TSK Fuzzy Systems
This is an incremental contribution providing a software tool for researchers and practitioners working with TSK fuzzy systems in machine learning.
The paper introduces PyTSK, a Python toolbox for developing Takagi-Sugeno-Kang (TSK) fuzzy systems, enabling optimization via fuzzy clustering or mini-batch gradient descent algorithms to improve generalization, particularly for big data applications.
This paper presents PyTSK, a Python toolbox for developing Takagi-Sugeno-Kang (TSK) fuzzy systems. Based on scikit-learn and PyTorch, PyTSK allows users to optimize TSK fuzzy systems using fuzzy clustering or mini-batch gradient descent (MBGD) based algorithms. Several state-of-the-art MBGD-based optimization algorithms are implemented in the toolbox, which can improve the generalization performance of TSK fuzzy systems, especially for big data applications. PyTSK can also be easily extended and customized for more complicated algorithms, such as modifying the structure of TSK fuzzy systems, developing more sophisticated training algorithms, and combining TSK fuzzy systems with neural networks. The code of PyTSK can be found at https://github.com/YuqiCui/pytsk.