Publications

Reinforced Dual-Flow Neural Network for Tabular Data Classification With Dynamical Transformer and Fuzzy Clustering

Online Feature Transformation Learning for Cross-Domain Object Category Recognition

In this paper, we introduce a new research problem termed online feature transformation learning in the context of multiclass object category recognition. The learning of a feature transformation is viewed as learning a global similarity metric function in an online manner. We first consider the problem of online learning a feature transformation matrix expressed in the original feature space and propose an online passive aggressive feature transformation algorithm. Then these original features are mapped to kernel space and an online single kernel feature transformation (OSKFT) algorithm is developed to learn a nonlinear feature transformation. Based on the OSKFT and the existing Hedge algorithm, a novel online multiple kernel feature transformation algorithm is also proposed, which can further improve the performance of online feature transformation learning in large-scale application. The classifier is trained with k nearest neighbor algorithm together with the learned similarity metric function. Finally, we experimentally examined the effect of setting different parameter values in the proposed algorithms and evaluate the model performance on several multiclass object recognition data sets. The experimental results demonstrate the validity and good performance of our methods on cross-domain and multiclass object recognition application.

2008 First International Conference on Complexity and Intelligence of the Artificial and Natural Complex Systems. Medical Applications of the Complex Systems.

Alternative Method of Constructing Granular Neural Networks

Utilizing granular computing to enhance artificial neural network architecture, a new type of network emerges-the granular neural network (GNN).GNNs offer distinct advantages over their traditional counterparts: The ability to process both numerical and granular data, leading to improved interpretability.This paper proposes a novel design method for constructing GNNs, drawing inspiration from existing interval-valued neural networks built upon NNNs.However, unlike the proposed algorithm in this work, which employs interval values or triangular fuzzy numbers for connections, existing methods rely on a pre-defined numerical network.This new method utilizes a uniform distribution of information granularity to granulate connections with unknown parameters, resulting in independent GNN structures.To quantify the granularity output of the network, the product of two common performance indices is adopted: The coverage of numerical data and the specificity of information granules.Optimizing this combined performance index helps determine the optimal parameters for the network.Finally, the paper presents the complete model construction and validates its feasibility through experiments on datasets from the UCI Machine Learning Repository.The results demonstrate the proposed algorithm's effectiveness and promising performance.

Evolutionary Optimization of Fuzzy Models

This study is devoted to the general identification methodology for fuzzy models. In contrast to numeric models, fuzzy models operate at a level of information granules — fuzzy sets and this brings up an important design requirement of transparency of the model. We propose a three-phase development framework by distinguishing between structural and parametric optimization processes. The underlying topology of the model dwells on fuzzy neural networks — architectures governed by fuzzy logic and equipped with parametric flexibility. Two general optimization mechanisms are explored: the structural optimization is realized via Genetic Programming whereas for the ensuing detailed parametric optimization we proceed with gradient-based learning. The main advantages of this approach are discussed in detail. The study is illustrated with the aid of a numeric example that provides a detailed insight into the performance of the fuzzy models.

ACDB-EA: Adaptive convergence-diversity balanced evolutionary algorithm for many-objective optimization

The theoretical fundamentals of learning theory based on fuzzy complex random samples

IEEE Transactions on Systems, Man, and Cybernetics publication information

is an organization within the framework of the IEEE, with professional interest in the closely interrelated fields of man-machine systems, systems science, systems engineering, and cybernetics.

Evolutionary design of hybrid self-organizing fuzzy polynomial neural networks with the aid of information granulation

ARIMA Feature-Based Approach to Time Series Classification

From Data to Information Granules: An Environment of Granular Computing

With enormous amounts of data come opportunities of building models of real-world systems that are instrumental in realizing a plethora of control, prediction, and classification tasks. The interpretability facet of ensuing models becomes highly relevant in light of designing autonomous systems and those constructs supporting human-centric decision-making environments. To transform data to tangible and actionable pieces of knowledge and formulate a problem at hand at a suitable level of abstraction, a convenient way to proceed is to position the problem in the environment of Granular Computing.We advocate that a systematic way of capturing knowledge residing within acquired data and encapsulating such knowledge in the form of interpretable models is supported by a suitable level of abstraction at which the data are to be represented. In turn, we show that an abstraction mechanism is conveniently realized in the form of information granules. Information granules and Granular Computing delivers an operational and flexible setting in which granular models are built and analyzed. A formal characterization of information granules is introduced where they are concisely described as triple (G, I, R) capturing their underlying geometry in the data space (G), information content (I), and representation capabilities of the underlying experimental evidence (R).A suite of design methods transforming data into information granules being articulated in various formal settings (e.g., intervals, fuzzy sets, rough sets) is analyzed and an array of generalizations is discussed (including collaborative ways of building granules in the presence of some auxiliary domain knowledge).In the sequel, it is shown how information granules regarded as functional modules are efficiently used in the construction of a vast array of interpretable models, especially rule-based architectures.

Adaptive Processing of Sequences and Data Structures,

Using multilayer perceptrons as receptive fields in the design of neural networks

Logic-driven autoencoders

The development of granular rule-based systems: a study in structural model compression

FedMPS: Federated Learning in a Synergy of Multi-Level Prototype-Based Contrastive Learning and Soft Label Generation

Federated learning (FL) facilitates collaborative training among multiple clients while preserving data privacy by eliminating raw data transmission. However, the inherent data heterogeneity among participants induces bias during collaborative learning, significantly degrading the performance of local models. Existing FL solutions face critical challenges in achieving efficient knowledge transmission, particularly with respect to insufficient information extraction or excessive communication costs, which result in slow convergence and inferior performance. To address these limitations, we propose a novel FL framework in a synergy of multi-level prototype-based contrastive learning (CL) and soft label generation, named FedMPS. The proposed method first constructs multi-level prototypes from different layers of the model to capture semantic information in high-level features and detailed information in low-level features. These prototypes are then utilized through CL to enhance intra-class discriminability and intra-class consistency in the feature space. In addition, a prototype-guided soft label generation module is introduced to model latent interclass relationships in the output space. Instead of exchanging model parameters, FedMPS transmits only prototypes and soft labels, effectively reducing global knowledge shift and communication costs. Extensive experimental studies on six publicly available datasets validate the effectiveness of the proposed method when compared to the current state-of-the-art FL approaches. The code is available at github.com/wenxinyang1026/FedMPS.

Classification with a limited space of features: Improving quality by rejecting misclassifications

Classification, especially in the case of a small space of features, is prone to errors. This is more important when it is costly to gain data from samples to calculate the values for futures. We study what effect limiting the space of features has on the performance of built classifiers and how the quality of classification can be improved by rejecting misclassified elements.

A design of information granule-based under-sampling method in imbalanced data classification