Selected work

Representative Papers

Supervised Spike Agreement Dependent Plasticity for Fast Local Learning in Spiking Neural Networks

Jan 13, 2026

This work proposes a supervised Synaptic Adaptation based on Discrepancy of Populations (SADP) learning rule that overcomes the limitations of traditional spike-timing-dependent plasticity (STDP), which relies on precise spike timing and pairwise updates and struggles to support efficient supervised learning. SADP introduces, for the first time, a population-level spike consistency metric—such as Cohen’s kappa—into the local synaptic update mechanism of spiking neural networks, eliminating the need for backpropagation, surrogate gradients, or teacher forcing. While preserving biological plausibility and hardware compatibility, the method enables efficient supervised training. Integrated with a hybrid CNN-SNN architecture and Poisson encoding, SADP achieves rapid convergence, strong performance, and remarkable hyperparameter robustness across MNIST, Fashion-MNIST, CIFAR-10, and biomedical image tasks, with a synaptic update mechanism exhibiting linear time complexity.

0 citationsRead paper

Spike Agreement Dependent Plasticity: A scalable Bio-Inspired learning paradigm for Spiking Neural Networks

Aug 22, 2025

To address the limitations of classical spike-timing-dependent plasticity (STDP)—namely, its reliance on precise spike timing and poor scalability—this paper proposes a biologically inspired Sequence Consistency-based Learning Rule (SC-LR). Instead of pairing individual pre- and postsynaptic spikes, SC-LR quantifies statistical consistency between population-level spike trains using Cohen’s kappa coefficient to measure neuronal co-activation. Synaptic dynamics are modeled via a spline kernel function derived from experimental data of organic memristive transistors. With an algorithmic complexity of O(N), SC-LR enables efficient hardware implementation. Evaluated on MNIST and Fashion-MNIST, SC-LR achieves 2.3–4.1% higher classification accuracy and 3.8× faster training than standard STDP, while preserving biological plausibility, learning efficacy, and computational scalability.

0 citationsRead paper

Fast Iterative and Task-Specific Imputation with Online Learning

Jan 23, 2025

To address the challenge of imputation under missing-not-at-random (MNAR) mechanisms, this paper proposes F3I—a task-aware, fast online imputation method. F3I integrates weighted k-nearest neighbors, iterative optimization, and online learning to enable end-to-end joint training of imputation and downstream classification/prediction tasks—an unprecedented capability in the literature. Theoretically, we establish a unified analytical framework that provides provable imputation quality guarantees across diverse missingness mechanisms, including MNAR. Empirically, F3I achieves state-of-the-art performance on synthetic benchmarks, drug repositioning, and handwritten digit recognition—outperforming mainstream imputation methods in both accuracy and computational efficiency. Its core innovations lie in (i) a task-driven online joint optimization paradigm and (ii) synergistic theoretical and practical guarantees specifically tailored for MNAR scenarios.

0 citationsRead paper

Recent publications

Latest Papers

Supervised Spike Agreement Dependent Plasticity for Fast Local Learning in Spiking Neural Networks

Jan 13, 2026

This work proposes a supervised Synaptic Adaptation based on Discrepancy of Populations (SADP) learning rule that overcomes the limitations of traditional spike-timing-dependent plasticity (STDP), which relies on precise spike timing and pairwise updates and struggles to support efficient supervised learning. SADP introduces, for the first time, a population-level spike consistency metric—such as Cohen’s kappa—into the local synaptic update mechanism of spiking neural networks, eliminating the need for backpropagation, surrogate gradients, or teacher forcing. While preserving biological plausibility and hardware compatibility, the method enables efficient supervised training. Integrated with a hybrid CNN-SNN architecture and Poisson encoding, SADP achieves rapid convergence, strong performance, and remarkable hyperparameter robustness across MNIST, Fashion-MNIST, CIFAR-10, and biomedical image tasks, with a synaptic update mechanism exhibiting linear time complexity.

0 citationsRead paper

Spike Agreement Dependent Plasticity: A scalable Bio-Inspired learning paradigm for Spiking Neural Networks

Aug 22, 2025

To address the limitations of classical spike-timing-dependent plasticity (STDP)—namely, its reliance on precise spike timing and poor scalability—this paper proposes a biologically inspired Sequence Consistency-based Learning Rule (SC-LR). Instead of pairing individual pre- and postsynaptic spikes, SC-LR quantifies statistical consistency between population-level spike trains using Cohen’s kappa coefficient to measure neuronal co-activation. Synaptic dynamics are modeled via a spline kernel function derived from experimental data of organic memristive transistors. With an algorithmic complexity of O(N), SC-LR enables efficient hardware implementation. Evaluated on MNIST and Fashion-MNIST, SC-LR achieves 2.3–4.1% higher classification accuracy and 3.8× faster training than standard STDP, while preserving biological plausibility, learning efficacy, and computational scalability.

0 citationsRead paper

Fast Iterative and Task-Specific Imputation with Online Learning

Jan 23, 2025

To address the challenge of imputation under missing-not-at-random (MNAR) mechanisms, this paper proposes F3I—a task-aware, fast online imputation method. F3I integrates weighted k-nearest neighbors, iterative optimization, and online learning to enable end-to-end joint training of imputation and downstream classification/prediction tasks—an unprecedented capability in the literature. Theoretically, we establish a unified analytical framework that provides provable imputation quality guarantees across diverse missingness mechanisms, including MNAR. Empirically, F3I achieves state-of-the-art performance on synthetic benchmarks, drug repositioning, and handwritten digit recognition—outperforming mainstream imputation methods in both accuracy and computational efficiency. Its core innovations lie in (i) a task-driven online joint optimization paradigm and (ii) synergistic theoretical and practical guarantees specifically tailored for MNAR scenarios.

0 citationsRead paper