Selected work

Representative Papers

Solution of the Hempel's statistical ambiguity problem and Causal AI

Jul 14, 2026

This work addresses Carl Hempel’s problem of statistical ambiguity—the challenge of deriving contradictory predictions from statistical regularities—by proposing a framework of Maximal Specific Causal Relationships (MSCRs) grounded in Nancy Cartwright’s probabilistic theory of causality. The approach formalizes causal rules, semantic probabilistic reasoning, and context-sensitive probability-raising models, integrating invariant feature learning with invariant causal prediction to systematically reconcile conflicting statistical information. The paper establishes, for the first time, a rigorous proof that MSCRs guarantee predictive consistency (Theorem 1), thereby demonstrating the solvability of the statistical ambiguity problem and offering a unified framework for causal artificial intelligence and causal machine learning that is both theoretically sound and computationally tractable.

0 citationsRead paper

Minimum distances of LDPC codes in 5G standard

Jul 06, 2026

This study addresses the minimum distance—a critical error-correction performance metric—of quasi-cyclic LDPC codes specified in the 5G NR standard, focusing on both high- and low-rate base graph 1 (BG1) configurations. By integrating algebraic analysis, combinatorial bounding algorithms, and cyclic modulo reduction techniques, the work establishes tight upper and lower bounds on the minimum distance for specific code instances: [9984, 8448] codes exhibit a minimum distance between 8 and 14, while [25344, 8448] codes range from 22 to 57. Furthermore, the paper introduces a novel early-termination strategy based on cyclic modulo reduction, which substantially reduces the computational complexity of parity-check operations during decoding. This approach enhances decoding efficiency without compromising error-correction performance.

0 citationsRead paper

Generalized Heavy-tailed Mutation for Evolutionary Algorithms

Apr 01, 2026

This work addresses the performance limitations of static mutation rates in OneMax-type problems by extending the theoretical foundation of heavy-tailed mutation within the (1+(λ,λ)) genetic algorithm framework. Specifically, it generalizes the underlying distribution from power-law to the broader class of regularly varying distributions and introduces a novel mutation operator satisfying this condition. The proposed approach maintains an expected optimization time of O(n) while overcoming the inherent constraints of fixed mutation rates. Theoretically, it outperforms any (1+(λ,λ)) algorithm employing a static mutation rate, and extensive experiments confirm its empirical efficacy.

0 citationsRead paper

Semi-Supervised 3D Segmentation for Type-B Aortic Dissection with Slim UNETR

Dec 19, 2025

High-quality annotations are scarce for 3D segmentation of Type B Aortic Dissection (TBAD), and existing methods lack robustness across multiple anatomical structures—true lumen (TL), false lumen (FL), and flap (FLT). Method: We propose a multi-output semi-supervised framework based on Slim UNETR, integrating multi-branch decoders, pseudo-labeling, and synergistic strong-weak data augmentation. Crucially, we introduce hypothesis-free probabilistic response consistency regularization—via rotation and flipping—for the first time in multi-output medical image segmentation, enabling end-to-end, post-processing-free semi-supervised training. Results: On the ImageTBAD dataset, our method achieves Dice scores of 89.7% (TL), 84.3% (FL), and 76.5% (FLT) using only 30% labeled data—surpassing both fully supervised baselines and state-of-the-art semi-supervised approaches. This demonstrates the efficacy of non-probabilistic consistency modeling for multi-structure segmentation in TBAD.

0 citationsRead paper

Mathematics of natural intelligence

Dec 07, 2025

This paper addresses the foundational problem of formalizing the high-order cognitive architecture—termed the *cognitome*—that emerges through biological evolution in the brain. To this end, it proposes the first unified mathematical framework: modeling the brain as a *neural hypernetwork*, with *COGs* (functional-system-and-cell assemblies) as elementary units, and formalizing consciousness as large-scale cognitive integration dynamics via *cognitive structural algebra*, *causal graph theory*, and *dynamical systems theory*. The framework rigorously derives, from first principles of causal discovery, natural taxonomies, prototype-based categorization, functional parcellation of the brain, and the integrated information theory of consciousness. It further unifies classical theories—including the global workspace, predictive coding, and the global neuronal workspace—through a single deductive pathway. Crucially, it demonstrates that higher cognitive functions strictly emerge from elementary causal inference mechanisms, thereby establishing a mathematically grounded foundation for natural intelligence.

0 citationsRead paper

Recent publications

Latest Papers

Solution of the Hempel's statistical ambiguity problem and Causal AI

Jul 14, 2026

This work addresses Carl Hempel’s problem of statistical ambiguity—the challenge of deriving contradictory predictions from statistical regularities—by proposing a framework of Maximal Specific Causal Relationships (MSCRs) grounded in Nancy Cartwright’s probabilistic theory of causality. The approach formalizes causal rules, semantic probabilistic reasoning, and context-sensitive probability-raising models, integrating invariant feature learning with invariant causal prediction to systematically reconcile conflicting statistical information. The paper establishes, for the first time, a rigorous proof that MSCRs guarantee predictive consistency (Theorem 1), thereby demonstrating the solvability of the statistical ambiguity problem and offering a unified framework for causal artificial intelligence and causal machine learning that is both theoretically sound and computationally tractable.

0 citationsRead paper

Minimum distances of LDPC codes in 5G standard

Jul 06, 2026

This study addresses the minimum distance—a critical error-correction performance metric—of quasi-cyclic LDPC codes specified in the 5G NR standard, focusing on both high- and low-rate base graph 1 (BG1) configurations. By integrating algebraic analysis, combinatorial bounding algorithms, and cyclic modulo reduction techniques, the work establishes tight upper and lower bounds on the minimum distance for specific code instances: [9984, 8448] codes exhibit a minimum distance between 8 and 14, while [25344, 8448] codes range from 22 to 57. Furthermore, the paper introduces a novel early-termination strategy based on cyclic modulo reduction, which substantially reduces the computational complexity of parity-check operations during decoding. This approach enhances decoding efficiency without compromising error-correction performance.

0 citationsRead paper

Generalized Heavy-tailed Mutation for Evolutionary Algorithms

Apr 01, 2026

This work addresses the performance limitations of static mutation rates in OneMax-type problems by extending the theoretical foundation of heavy-tailed mutation within the (1+(λ,λ)) genetic algorithm framework. Specifically, it generalizes the underlying distribution from power-law to the broader class of regularly varying distributions and introduces a novel mutation operator satisfying this condition. The proposed approach maintains an expected optimization time of O(n) while overcoming the inherent constraints of fixed mutation rates. Theoretically, it outperforms any (1+(λ,λ)) algorithm employing a static mutation rate, and extensive experiments confirm its empirical efficacy.

0 citationsRead paper

Semi-Supervised 3D Segmentation for Type-B Aortic Dissection with Slim UNETR

Dec 19, 2025

High-quality annotations are scarce for 3D segmentation of Type B Aortic Dissection (TBAD), and existing methods lack robustness across multiple anatomical structures—true lumen (TL), false lumen (FL), and flap (FLT). Method: We propose a multi-output semi-supervised framework based on Slim UNETR, integrating multi-branch decoders, pseudo-labeling, and synergistic strong-weak data augmentation. Crucially, we introduce hypothesis-free probabilistic response consistency regularization—via rotation and flipping—for the first time in multi-output medical image segmentation, enabling end-to-end, post-processing-free semi-supervised training. Results: On the ImageTBAD dataset, our method achieves Dice scores of 89.7% (TL), 84.3% (FL), and 76.5% (FLT) using only 30% labeled data—surpassing both fully supervised baselines and state-of-the-art semi-supervised approaches. This demonstrates the efficacy of non-probabilistic consistency modeling for multi-structure segmentation in TBAD.

0 citationsRead paper

Mathematics of natural intelligence

Dec 07, 2025

This paper addresses the foundational problem of formalizing the high-order cognitive architecture—termed the *cognitome*—that emerges through biological evolution in the brain. To this end, it proposes the first unified mathematical framework: modeling the brain as a *neural hypernetwork*, with *COGs* (functional-system-and-cell assemblies) as elementary units, and formalizing consciousness as large-scale cognitive integration dynamics via *cognitive structural algebra*, *causal graph theory*, and *dynamical systems theory*. The framework rigorously derives, from first principles of causal discovery, natural taxonomies, prototype-based categorization, functional parcellation of the brain, and the integrated information theory of consciousness. It further unifies classical theories—including the global workspace, predictive coding, and the global neuronal workspace—through a single deductive pathway. Crucially, it demonstrates that higher cognitive functions strictly emerge from elementary causal inference mechanisms, thereby establishing a mathematically grounded foundation for natural intelligence.

0 citationsRead paper