Selected work

Representative Papers

Limits of Kernelization and Parametrization for Phylogenetic Diversity with Dependencies

Feb 13, 2026

In the Maximize Phylogenetic Diversity problem, we are given a phylogenetic tree that represents the genetic proximity of species, and we are asked to select a subset of species of maximum phylogenetic diversity to be preserved through conservation efforts, subject to budgetary constraints that allow only k species to be saved. This neglects that it is futile to preserve a predatory species if we do not also preserve at least a subset of the prey it feeds on. Thus, in the Optimizing PD with Dependencies ($\epsilon$-PDD) problem, we are additionally given a food web that represents the predator-prey relationships between species. The goal is to save a set of k species of maximum phylogenetic diversity such that for every saved species, at least one of its prey is also saved. This problem is NP-hard even when the phylogenetic tree is a star. The $\alpha$-PDD problem alters PDD by requiring that at least some fraction $\alpha$ of the prey of every saved species are also saved. In this paper, we study the parameterized complexity of $\alpha$-PDD. We prove that the problem is W[1]-hard and in XP when parameterized by the solution size k, the diversity threshold D, or their complements. When parameterized by the vertex cover number of the food web, $\alpha$-PDD is fixed-parameter tractable (FPT). A key measure of the computational difficulty of a problem that is FPT is the size of the smallest kernel that can be obtained. We prove that, when parameterized by the distance to clique, 1-PDD admits a linear kernel. Our main contribution is to prove that $\alpha$-PDD does not admit a polynomial kernel when parameterized by the vertex cover number plus the diversity threshold D, even if the phylogenetic tree is a star. This implies the non-existence of a polynomial kernel for $\alpha$-PDD also when parameterized by a range of structural parameters of the food web, such as its dist[...]

3 citationsRead paper

FSM Modeling For Off-Blockchain Computation

Jun 02, 2025

To address the high on-chain execution overhead and inflexible layering of smart contracts, this paper proposes a blockchain-onchain/offchain cooperative computing framework based on finite state machines (FSMs). First, it integrates separation-of-concerns principles with FSM modeling to enable patterned decomposition of contract logic and automatic identification of migratable code fragments. Second, it formally defines offchain migration patterns—including their syntactic structure and precise triggering conditions—for the first time. Third, it employs model-driven development (MDD) to automatically generate bidirectional interaction interfaces that guarantee state consistency and security. Evaluated on the Ethereum testnet, the framework reduces gas consumption by 37%, significantly alleviating on-chain computational load. This work establishes the first systematic, formally verifiable methodology for modeling and generating lightweight smart contract executions, supporting rigorous offchain migration while preserving correctness and trustworthiness.

1 citationsRead paper

Accelerating Neural Network Training: An Analysis of the AlgoPerf Competition

Feb 20, 2025

This study addresses the fundamental question: “Can purely algorithmic improvements yield practical acceleration in neural network training?” To this end, we organized the inaugural AlgoPerf competition, establishing— for the first time—two rigorous evaluation paradigms: workload-agnostic assessment and hyperparameter-free benchmarking, with end-to-end training time on identical hardware as the sole primary metric. Methodologically, we developed a multi-task benchmarking framework integrating Distributed Shampoo (a non-diagonal preconditioner) and Schedule-Free AdamW (a hyperparameter-free optimizer), complemented by standardized temporal measurement protocols and fairness-preserving engineering safeguards. Results show that Distributed Shampoo achieved top performance in the hyperparameter-tuned track, while Schedule-Free AdamW led in the hyperparameter-free track. Top-performing methods demonstrated consistent speedups across diverse CV and NLP tasks, empirically validating that high-quality algorithmic design delivers substantial and robust training acceleration.

1 citationsRead paper

Dynamical modelling of the frailty index indicates that health reaches a tipping point near age 75

Dec 02, 2024

Human healthy aging exhibits a critical transition point whose timing and mechanistic basis remain poorly understood, particularly regarding the abrupt increase in mortality risk and sex-specific differences in frailty progression. Method: Leveraging >250,000 longitudinal health-state transitions from population-based cohorts, we developed a dynamic model quantifying imbalance between damage accumulation and repair in the frailty index (FI) trajectory. Contribution/Results: We identify ~75 years as a systemic critical transition: prior to this age, health status remains relatively stable; thereafter, FI acceleration reflects equilibration of damage and repair rates, coinciding with synchronous decline in health robustness and resilience. This critical point mechanistically explains the sharp rise in mortality and underpins the “frailty paradox”—the counterintuitive observation that women exhibit higher frailty yet lower mortality—through sex-dimorphic deterioration patterns. Findings are robustly replicated across independent longitudinal cohorts (HRS, ELSA) and consistently observed in both sexes.

1 citationsRead paper

Smart Contracts for SMEs and Large Companies

Apr 02, 20242024 IEEE Virtual Conference on Communications (VCC)

Significant disparities in IT capabilities across enterprises—from SMEs to large corporations—and the high barrier to smart contract development (requiring expertise in low-level languages like Solidity) hinder widespread adoption of blockchain technology. Method: This paper proposes a BPMN-centric framework for automated smart contract generation, directly translating business process models into blockchain smart contracts that support multi-step transactions and hot upgrades—without requiring users to possess blockchain or programming knowledge. Contribution/Results: The approach introduces the first zero-code, non-blockchain-expert solution for generating maintainable, debuggable, and business-semantic-preserving smart contracts. It features a lightweight BPMN-to-Contract compilation mechanism and a dynamic upgrade protocol. Experimental evaluation demonstrates feasibility in cross-scale enterprise collaboration scenarios, reducing smart contract development time by over 70%. The framework significantly advances the democratization and industrial deployment of smart contract technology.

1 citationsRead paper

Recent publications

Latest Papers

Does Latent Context Help? A Controlled Evaluation of Inverse Reinforcement Learning in Arctic Shipping

Aug 06, 2026

This study addresses the need for explainability and environmental robustness in AI-assisted navigation for Arctic shipping by proposing a reward modeling approach based on inverse reinforcement learning (IRL). The authors systematically compare linear shared reward models, nonlinear shared reward models, and a meta-IRL model incorporating vessel-class latent variables, using preregistered feature-masking ablation experiments and multidimensional evaluation metrics—including predictive accuracy, route fidelity, and reward transferability. Results show that the nonlinear reward model improves held-out likelihood by 50.9% over the linear baseline, while introducing latent variables degrades performance by 16.5%. This indicates that heterogeneity in vessel behavior is primarily driven by observable route and environmental factors rather than unobserved preferences.

0 citationsRead paper

Transferable Dual-Stream Representations for Mesoscale-Preserving Sea Surface Temperature Downscaling

Aug 04, 2026

Existing sea surface temperature (SST) downscaling methods often fail to preserve mesoscale eddies—features critical to ocean dynamics—resulting in overly smoothed outputs, spectral distortions, and poor generalization. To address this, this work proposes EddyFlow, a novel framework that introduces, for the first time, a physics-informed dual-stream representation learning mechanism to jointly optimize prediction accuracy, structural fidelity, and cross-regional transferability at kilometer-scale resolution. By integrating deep representation learning, multiscale spectral analysis, physics-constrained loss functions, and cross-domain strategies, EddyFlow achieves substantial performance gains under zero-shot and few-shot settings: it reduces zero-shot RMSE by 21% over unseen ocean regions, attains a relative persistence skill of 85.6%, and yields a power spectral density ratio approaching the ideal value of 1.00.

0 citationsRead paper

Designing a digital word-learning intervention with neurodiverse children: Experiences and ideas from children with developmental language disorder

Jul 31, 2026

Developmental Language Disorder (DLD) affects approximately 7% of children, leading to significant difficulties in vocabulary acquisition and associated social and academic challenges, yet this population has been largely overlooked in participatory design research. This study pioneers the application of a tailored co-design approach with children with DLD, engaging them in the design of digital vocabulary interventions through low-cognitive-load activities such as storytelling characters and video games. Drawing on data from drawings, sentence-level feedback, field notes, and emotion scales, reflexive thematic analysis revealed key design preferences and needs. Findings indicate a strong preference for familiar characters and gamified elements, and underscore the necessity of integrating progress indicators, multimodal presentation, and real-world adult support into intervention frameworks. The study demonstrates the feasibility and effectiveness of customized co-design methodologies for children with DLD.

0 citationsRead paper

Uncertainty quantification for trustworthy deep learning: Methods and measures

Jul 30, 2026

This work addresses the challenge of unreliable predictive uncertainty estimation in deep neural networks, which undermines their trustworthiness in safety-critical applications. The paper presents a systematic survey of uncertainty quantification methods, with a focus on ensemble and approximate Bayesian techniques, and introduces a decoupled “method–metric” framework that unifies the generation of predictive distributions and the aggregation of uncertainties. By integrating diverse approaches—including Bayesian neural networks, Monte Carlo Dropout, deep and efficient ensembles, single-forward methods, evidential networks, conformal prediction, and post-hoc calibration—the study establishes a unified taxonomy and evaluation benchmark. This enables a clear delineation of each method’s theoretical foundations, implementation strategies, empirical performance, and limitations, while also outlining promising directions for uncertainty research in large language models.

0 citationsRead paper

Covariance Last-Layer Ensembles: Function-Space Diversity for Efficient Uncertainty Quantification

Jul 26, 2026

This work addresses the limited functional diversity in Last-Layer Ensemble (LLE) methods—caused by shared backbone gradients—which degrades out-of-distribution (OOD) detection performance. To overcome this, the authors propose Covariance Last-Layer Ensemble (cov-LLE), which directly imposes a covariance penalty in function space to explicitly enhance ensemble member diversity, surpassing the indirect decorrelation achieved by conventional weight orthogonality. The method introduces a scale-invariant, label-free directional uncertainty score that substantially improves near-OOD detection. Without increasing forward-pass computational cost or compromising in-distribution accuracy, cov-LLE boosts predictive variance from 0.05 to 9.3 (approaching the 22.1 of deep ensembles), reduces expected calibration error (ECE) from 0.135 to 0.090, and consistently improves ROC AUC by 0.16–0.18 across diverse backbone architectures.

0 citationsRead paper