Selected work

Representative Papers

Reservoir computing and photoelectrochemical sensors: A marriage of convenience

Jul 01, 2023Coordination chemistry reviews

To address the need for real-time, low-power detection of complex chemical components in biological fluids and environmental samples, existing photoelectrochemical (PEC) sensors face critical bottlenecks—including reliance on energy-intensive digital hardware for signal processing and insufficient robustness. This work introduces, for the first time, physical reservoir computing (PRC) into PEC sensing systems, leveraging the intrinsic nonlinear dynamics of the sensor itself as a natural analog computational resource to enable event-driven, in-situ information processing. By eliminating conventional digital signal processing modules, the approach drastically reduces power consumption and latency. In dynamic detection tasks for glucose and dopamine, the system achieves millisecond-scale response times and 98.2% classification accuracy, while reducing power consumption by two orders of magnitude compared to standard approaches. This work establishes a novel paradigm for neuromorphic sensing, brain-inspired molecular perception, and edge-intelligent chemical sensing.

14 citationsRead paper

Embedded Graph Convolutional Networks for Real-Time Event Data Processing on SoC FPGAs

Jun 11, 2024arXiv.org

To address the challenges of high-throughput, ultra-low-latency, and energy-efficient real-time event processing for automotive embedded systems, this paper proposes a hardware-software co-optimization framework targeting SoC FPGAs. We present the first PointNet++ acceleration implementation on the Xilinx ZCU104 platform and introduce an event-aware asynchronous graph convolutional network (EFGCN) capable of online analysis of continuous event streams. Our approach integrates model pruning, quantization, and a customized pipelined accelerator architecture, achieving over 100× model size reduction. Experimental evaluation demonstrates a throughput of 13.3 MEPS and an end-to-end latency of 4.47 ms, with only 2.3% and 1.7% accuracy degradation on N-Caltech101 and N-Cars benchmarks, respectively. This work establishes the first hardware architecture for asynchronous GCNs, and we publicly release the complete software-hardware stack. Our framework provides an efficient, edge-deployable paradigm for event-driven intelligent perception.

6 citations1 influentialRead paper

KNOWM memristors in a bridge synapse delay-based reservoir computing system for detection of epileptic seizures

Jun 26, 2022Int. J. Parallel Emergent Distributed Syst.

To address the need for real-time, low-power epileptic seizure detection, this work proposes a novel memristor-based bridge synaptic delayed reservoir computing architecture using KNOWM memristors. The method integrates four KNOWM memristors with a differential amplifier to construct a single-node echo state machine (SNESM) as a physical reservoir, leveraging feedback loops to perform nonlinear temporal transformation and disentangle complexity features from raw EEG signals. This represents the first hardware implementation of a memristive bridge synaptic structure for delayed reservoir computing, significantly enhancing discriminability among complexity metrics across distinct epileptic states. Experimental results demonstrate reduced inter-feature correlation and improved class separability in the transformed signal space compared to raw EEG, leading to markedly higher seizure detection accuracy. The approach establishes a new paradigm for neuromorphic edge-intelligent healthcare systems.

6 citations1 influentialRead paper

Balancing the Spread of Two Opinions in Sparse Social Networks

May 21, 2021AAAI Conference on Artificial Intelligence

This paper addresses the balance control problem of dual-opinion co-propagation in sparse social networks: given a budget, propagation rounds, and an initial seed set, minimize seed expansion such that every node ultimately holds either zero or both opinions—achieving global opinion balance. We innovatively embed a dual-threshold adoption mechanism into a target-set selection framework, proposing a discrete propagation model that jointly captures single- and dual-opinion adoption tendencies. Theoretically, we prove the problem is fixed-parameter tractable (FPT) with respect to the vertex cover number and devise an efficient parameterized algorithm. Moreover, we establish its polynomial-time solvability on sparse graph classes—including trees and degenerate graphs. Our work provides the first parameterized solution for multi-opinion dynamic control in sparse networks, backed by rigorous theoretical guarantees.

6 citationsRead paper

Participatory Budgeting Project Strength via Candidate Control

May 28, 2025Adaptive Agents and Multi-Agent Systems

This study investigates the computational complexity of manipulating participatory budgeting elections by adding or deleting candidate projects to either ensure a target project’s selection (constructive control) or its exclusion (destructive control). It presents the first systematic analysis of candidate control under prominent voting rules—Phragmén, Equal Shares, and GreedyAV—and introduces a novel perspective that evaluates project strength through the lens of candidate deletions. The theoretical findings reveal that the problem is NP-hard under most rules, yet polynomial-time algorithms exist for GreedyAV and in the unit-cost setting. Experimental results corroborate the efficacy of the proposed approach, offering both theoretical insights and practical tools for assessing the robustness of participatory budgeting outcomes and the relative importance of individual projects.

2 citationsRead paper

Recent publications

Latest Papers

Fine-tuned Normalizing Flows for ALICE Zero Degree Calorimeter Fast Simulation

Aug 13, 2026

This work addresses the high computational cost and low efficiency of traditional Monte Carlo methods in simulating neutron detector responses in the ALICE Zero Degree Calorimeter. To overcome these limitations, the authors propose a generative surrogate model based on normalizing flows, enhanced with transfer learning and conditional fine-tuning. They introduce physics-informed evaluation metrics—including conditional weighted MAE, dispersion ratio, and Jaccard co-activation error—and devise a progressive unfreezing fine-tuning strategy to accurately capture the input–output physical dependencies. Experimental results demonstrate that the proposed model achieves a Wasserstein distance of 1.61 ± 0.02, significantly outperforming baseline approaches while simultaneously improving both fidelity and inference speed.

0 citationsRead paper

Derivative Computation in PINNs: Automatic Differentiation, Finite Differences and Beyond

Aug 11, 2026

This work addresses the high computational cost and memory consumption of automatic differentiation (AD) in physics-informed neural networks (PINNs), as well as its susceptibility to silent errors in architectures involving inter-sample dependencies such as BatchNorm or self-attention. The study presents the first systematic evaluation of finite differences (FD) as an alternative for derivative computation in PINNs, introducing a calibrated step-size strategy and a stochastic FD variant. It further proposes a sample-wise gradient approximation method that requires only forward passes. Experiments on three benchmark partial differential equations demonstrate that, in full-batch settings, FD achieves comparable accuracy to AD while being faster and using less memory. The proposed stochastic FD excels particularly in steady-state problems, and crucially, FD yields derivative errors an order of magnitude lower than AD in models with inter-sample dependencies.

0 citationsRead paper

One Model to Magnify Them All: Efficient Scale-Invariant Histopathology via Conditional Normalization and Continuous Magnification Training

Aug 10, 2026

This work addresses the challenge that whole-slide images (WSIs) in digital pathology are acquired across continuously varying magnifications, while existing deep learning models are scale-sensitive and struggle to generalize to unseen or misaligned magnification levels. To overcome this limitation, the authors propose Conditional Layer Normalization (CLN), a lightweight mechanism that employs a small MLP to dynamically generate normalization parameters based on the input pixel size. Integrated into standard CNN architectures and trained on image patches sampled across a continuous range of scales, CLN enables a single model to achieve strong generalization across arbitrary magnifications. Notably, this approach is the first to cover a continuous spectrum of magnifications without requiring ensemble models. On the PANDA prostate cancer dataset, it matches or exceeds the performance of dedicated single-magnification models, consistently ranking among the top three across all evaluated magnifications—including unseen ones—while reducing both training and inference costs by 4–5×.

0 citationsRead paper

A systematic framework for the identification and statistical quantification of impulsivity in condition monitoring signals

Aug 08, 2026

This study addresses the challenge in industrial machinery condition monitoring where impulsive interferences are often confounded with genuine fault signatures, thereby degrading diagnostic accuracy. To resolve this issue, the authors propose a two-stage data-driven framework: first, the Mann-Whitney U statistic is employed to objectively select the optimal impulsiveness metric; second, bootstrap resampling combined with Monte Carlo simulation is used to assess the statistical significance and quantify the intensity of detected impulses. This approach uniquely integrates impulsiveness metric selection and significance testing into a scalable, standardized pipeline, substantially enhancing diagnostic robustness. Experimental validation on three synthetic signal types and real-world vibration data from an industrial compressor demonstrates that the proposed framework effectively discriminates between normal operation, localized damage, and anomalous interference.

0 citationsRead paper

Disentangling Language Modeling and Boundaries

Aug 04, 2026

This work addresses the challenge that existing language models, constrained by model-specific tokenizers, struggle to efficiently transfer knowledge or flexibly adjust segmentation boundaries across architectures. The authors propose decoupling the “next-byte prediction” and “segment boundary delineation” distributions within byte-level language models, leveraging their shared output space to enable precise, alignment-free knowledge transfer. Through distributional decoupling analysis, cross-model capability transfer experiments, and boundary behavior measurements, the study provides the first systematic evidence that modeling capacity and boundary decisions can be approximately controlled independently. These findings offer preliminary support for the decoupling hypothesis and lay the groundwork for a universal model ecosystem standardized on byte-level interfaces, potentially enabling low-cost, routine model capability transfer and boundary reconfiguration.

0 citationsRead paper