Abstract:Pattern recognition plays an important role in the process of knowledge discovery. The construction of easily describable and interpretable classification rules is of vital importance in pattern recognition. In this study, we propose a development of fuzzy rule-based classifier for multiclass classification problems and elaborate on a privacy-preserving realization of the proposed methodology in the presence of decentralized datasets. Fuzzy rule-based models provide an effective and efficient alternative for characterizing the complex relationship between the input variables and target classes. An overall design process of the proposed classifier consists of two main phases: (a) formation of information granules (clusters) to reveal the underlying structure of the training data, and (b) construction of local classification rules whose outputs reflect the probability distribution of the input data over all the classes. The constructed information granules form a backbone of the architecture of the classifier while the optimization of the parameters of local rules is carried out through using a gradient descent method with the guidance of the cross-entropy loss function. Furthermore, a federated gradient-based optimization mechanism is utilized to construct fuzzy classifier in a privacy-preserving approach. The originalities of the proposed methodology are twofold: first, a design of fuzzy classifier through the synergy of cluster-centric architecture and the cross-entropy loss function is presented. Second, we augment the proposed fuzzy classifier based on the concept of federated learning such that it can learn from distributed data without sacrificing data security and confidentiality. Experiments are carried out on a two-dimensional synthetic dataset and a number of real-world datasets. Experimental results show the excellent classification capability of the proposed classifier realized in the centralized way and in the federated learning environment.

Horizontal Federated Learning of Takagi–Sugeno Fuzzy Rule-Based Models

Privacy-Preserving Realization of Fuzzy Clustering and Fuzzy Modeling Through Vertical Federated Learning

A Design of Fuzzy Rule-based Classifier for Multi-class Classification and Its Realization in Horizontal Federated Learning

Privacy-Preserving and Reliable Distributed Federated Learning

From distributed machine learning to federated learning: In the view of data privacy and security

Privacy, accuracy, and model fairness trade-offs in federated learning

Concurrent vertical and horizontal federated learning with fuzzy cognitive maps

A Communication-efficient Federated Learning Assisted by Central Data: Implementation of Vertical Training into Horizontal Federated Learning

A Privacy-Enhanced Federated Learning Model Training Method

Federated FCM: Clustering under Privacy Requirements

A Lightweight and Accuracy-Lossless Privacy-Preserving Method in Federated Learning

Privacy-Preserving Federated Learning in Fog Computing

Secure Federated Submodel Learning

Privacy preserving federated learning for full heterogeneity

How global observation works in Federated Learning: Integrating vertical training into Horizontal Federated Learning

Prioritized Multi-Criteria Federated Learning

Fairness and Accuracy in Horizontal Federated Learning

LF3PFL: A Practical Privacy-Preserving Federated Learning Algorithm Based on Local Federalization Scheme

Enhancing Model Performance Via Vertical Federated Learning for Non-Overlapping Data Utilization

FedPH: Privacy-enhanced Heterogeneous Federated Learning

A Study of Federated Learning with Internet of Things for Data Privacy and Security using Privacy Preserving Techniques