Selected work

Representative Papers

WIRE: Write Energy Reduction via Encoding in Phase Change Main Memories (PCM)

Nov 07, 2025

Phase-change memory (PCM) suffers from high write energy consumption and limited endurance. To address these challenges, this paper proposes a low-overhead encoding scheme: a dynamic frequent-value-stack-based encoding that ensures most writes induce only a single-bit flip; combined with block-level wear leveling and differential bit rotation, it achieves effective write distribution. The core techniques include frequent value identification, Hamming distance optimization, bit-flip minimization, and efficient encoding storage. Experimental evaluation under multithreaded and multiprogrammed workloads demonstrates that the proposed approach reduces write energy by 38.2% on average, decreases bit flips by 41.7%, and extends PCM lifetime by 2.3×, significantly improving main memory energy efficiency and reliability.

1 citationsRead paper

SMART-WRITE: Adaptive Learning-Based Write Energy Optimization for Phase Change Memory

Jan 06, 2025Computing and Communication Workshop and Conference

Phase-change memory (PCM) faces two critical bottlenecks—high write energy consumption and limited cell endurance—that hinder its adoption as a DRAM replacement. To address these challenges, this work proposes an adaptive learning–based online write optimization framework that jointly leverages a lightweight neural network for real-time sensing of device state and aging characteristics, and deep reinforcement learning for dynamic selection of optimal write parameters. The framework ensures data reliability while simultaneously optimizing energy efficiency and write performance. Experimental evaluation demonstrates that, compared to baseline approaches, the proposed method achieves an average 63% reduction in write energy, up to a 51% improvement in write throughput, and significantly extends PCM device lifetime. This work establishes a deployable intelligent control paradigm for high-efficiency, long-endurance PCM systems.

1 citationsRead paper

Moments of crosscorrelation demerit factors of binary sequences

Sep 04, 2026

Families of sequences with low mutual aperiodic crosscorrelation assist the design of systems for multi-user asynchronous communications and multiple-input multiple-output radar. The crosscorrelation demerit factor of a pair of sequences is the sum of the squared magnitudes of their crosscorrelation values at every shift when the sequences are normalized to unit Euclidean norm, and the merit factor is the reciprocal of the demerit factor. For each positive integer $\ell$, we endow the $2^{2 \ell}$ pairs of binary sequences of length $\ell$ with uniform probability measure and study the distribution of their crosscorrelation demerit factors. Sarwate showed that the mean value is always $1$ regardless of length $\ell$. We develop a method for finding an exact formula for the $p$th central moment (for any positive integer $p$) as a function of $\ell$. Formulae for the variance and third central moment ($p=2$ and $3$) are then obtained by hand calculations, while the fourth through sixth central moments are obtained by computer-assisted calculations. Our theory also shows that all the central moments must be strictly positive for $p\geq 2$ and $\ell \geq 3$.

0 citationsRead paper

Adaptive Gated Deepfake Detection for Low-Resolution and Resource-Constrained Environments

Sep 04, 2026

本文提出AdaGate-DF框架，通过图像质量线索在低分辨率和资源受限环境中自适应地进行深度伪造检测，以提高效率并保持高准确率。

0 citationsRead paper

TacStyle: Personalizing Tactile Robot Policies using Structured Behavior Representations

Jun 12, 2026

Current robotic systems struggle to accurately interpret user preferences regarding operational styles—such as applied force—expressed through natural language. To address this challenge, this work proposes a novel approach that constructs a structured, continuous, and interpretable latent action space. Instead of directly generating behaviors from language, foundation models map linguistic preference prompts into this latent space, enabling the system to reason about and generate tactile manipulation policies aligned with user expectations. By integrating structured representation learning with language-guided preference inference, the method achieves fine-grained, high-fidelity personalization of robot behavior. Experiments in both simulation and real-world settings demonstrate its effectiveness, showing that only a small number of preference labels are sufficient to adapt behaviors precisely to individual user preferences.

0 citationsRead paper

Recent publications

Latest Papers

Moments of crosscorrelation demerit factors of binary sequences

Sep 04, 2026

Families of sequences with low mutual aperiodic crosscorrelation assist the design of systems for multi-user asynchronous communications and multiple-input multiple-output radar. The crosscorrelation demerit factor of a pair of sequences is the sum of the squared magnitudes of their crosscorrelation values at every shift when the sequences are normalized to unit Euclidean norm, and the merit factor is the reciprocal of the demerit factor. For each positive integer $\ell$, we endow the $2^{2 \ell}$ pairs of binary sequences of length $\ell$ with uniform probability measure and study the distribution of their crosscorrelation demerit factors. Sarwate showed that the mean value is always $1$ regardless of length $\ell$. We develop a method for finding an exact formula for the $p$th central moment (for any positive integer $p$) as a function of $\ell$. Formulae for the variance and third central moment ($p=2$ and $3$) are then obtained by hand calculations, while the fourth through sixth central moments are obtained by computer-assisted calculations. Our theory also shows that all the central moments must be strictly positive for $p\geq 2$ and $\ell \geq 3$.

0 citationsRead paper

Adaptive Gated Deepfake Detection for Low-Resolution and Resource-Constrained Environments

Sep 04, 2026

本文提出AdaGate-DF框架，通过图像质量线索在低分辨率和资源受限环境中自适应地进行深度伪造检测，以提高效率并保持高准确率。

0 citationsRead paper

TacStyle: Personalizing Tactile Robot Policies using Structured Behavior Representations

Jun 12, 2026

Current robotic systems struggle to accurately interpret user preferences regarding operational styles—such as applied force—expressed through natural language. To address this challenge, this work proposes a novel approach that constructs a structured, continuous, and interpretable latent action space. Instead of directly generating behaviors from language, foundation models map linguistic preference prompts into this latent space, enabling the system to reason about and generate tactile manipulation policies aligned with user expectations. By integrating structured representation learning with language-guided preference inference, the method achieves fine-grained, high-fidelity personalization of robot behavior. Experiments in both simulation and real-world settings demonstrate its effectiveness, showing that only a small number of preference labels are sufficient to adapt behaviors precisely to individual user preferences.

0 citationsRead paper

Euclidean Steiner Shallow-Light Trees in Higher Dimensions

May 26, 2026

This study addresses the construction of Steiner shallow-light trees (SLTs) in high-dimensional Euclidean spaces, aiming to simultaneously achieve near-shortest paths and low lightness—both independent of the ambient dimension. For any finite point set, designated root, and accuracy parameter ε, the authors propose a method grounded in geometric graph theory and approximation algorithms. By introducing Steiner points and reanalyzing the planar core structure along with its high-dimensional generalization, they construct an SLT with stretch factor (1+ε) and lightness O(√(1/ε)). This result provides the first dimension-independent performance guarantee for SLTs in arbitrary dimensions, thereby affirming Solomon’s conjecture on the existence of such structures in high-dimensional Euclidean spaces.

0 citationsRead paper

Consistent Variable Selection for GARCH-X Models

Apr 28, 2026

This study addresses the challenge of identifying truly relevant exogenous covariates in GARCH-X models for volatility dynamics. The authors propose a variable selection method based on multiple hypothesis testing, which integrates a Wald-type test statistic with the Benjamini–Yekutieli false discovery rate (FDR) control procedure. They establish, for the first time, an asymptotically consistent variable selection rule under the GARCH-X framework that rigorously controls the FDR. Monte Carlo simulations demonstrate that the proposed method exhibits strong accuracy and robustness across various error distributions and dependence structures. Empirical application to S&P 500 index volatility modeling further confirms its practical effectiveness.

0 citationsRead paper