Selected work

Representative Papers

Causal explanations of outliers in systems with lagged time-dependencies

Feb 04, 2026

This work addresses the challenge of root cause localization in time-varying dynamic systems with lag effects and memory properties—such as energy systems—where anomalies exhibit complex temporal dependencies. For the first time, the authors extend a strictly causal root cause analysis framework to such systems, introducing two truncation strategies to manage infinite-time dependency graphs: one preserving the original causal mechanisms and the other employing mechanism approximation. By integrating causal graph modeling with a data generation approach tailored to energy consumption peak scenarios, the proposed method is evaluated in a simulated factory environment. Results demonstrate that, given sufficient lag order, the approach accurately identifies the spatiotemporal origins of anomalies, while also quantifying the performance trade-offs introduced by mechanism approximation.

0 citationsRead paper

VCO-CARE: VCO-based Calibration-free Analog Readout for Electrodermal activity sensing

Sep 08, 2025

To address the challenges of high sensitivity, ultra-low power consumption, and calibration-free operation in skin electrodermal activity (EDA) sensing for wearable devices, this work proposes a calibration-free analog front-end (AFE) architecture based on a voltage-controlled oscillator (VCO). Instead of conventional resistance or current measurement paradigms requiring subject-specific calibration, the design employs frequency-domain readout to directly map conductance variations. It achieves 40 ps resolution and <0.0025% relative error over a 0–1.5 Hz bandwidth. Through low-noise circuit design and post-layout optimization, the system attains an average power consumption of only 2.3 μW and an input-referred noise of 0.8 μV_rms. The proposed AFE significantly mitigates inter-subject variability, enabling robust, ultra-low-power, on-chip EDA monitoring suitable for continuous, multi-user applications.

0 citationsRead paper

Effect Identification and Unit Categorization in the Multi-Score Regression Discontinuity Design with Application to LED Manufacturing

Aug 21, 2025

Real-world decision-making often relies on multi-dimensional threshold rules, yet conventional multi-score regression discontinuity designs (MRD) collapse these into a unidimensional framework, inducing noncompliance and biasing causal effect estimation. This paper establishes, for the first time, a classification framework for unit-level behavior under multi-dimensional cutoffs—explicitly defining compliers, avoiders, and inert agents. It introduces identification conditions for local average treatment effects at the sub-rule level, uncovering how rule decomposition shapes behavioral responses. We develop a novel MRD estimator leveraging multiple scores and validate it using both simulation studies and real-world semiconductor manufacturing data. Relative to standard MRD, our approach substantially reduces estimation variance and improves policy evaluation accuracy. Empirical application to LED production line optimization demonstrates its practical utility in industrial policy design.

0 citationsRead paper

Recent publications

Latest Papers

Causal explanations of outliers in systems with lagged time-dependencies

Feb 04, 2026

This work addresses the challenge of root cause localization in time-varying dynamic systems with lag effects and memory properties—such as energy systems—where anomalies exhibit complex temporal dependencies. For the first time, the authors extend a strictly causal root cause analysis framework to such systems, introducing two truncation strategies to manage infinite-time dependency graphs: one preserving the original causal mechanisms and the other employing mechanism approximation. By integrating causal graph modeling with a data generation approach tailored to energy consumption peak scenarios, the proposed method is evaluated in a simulated factory environment. Results demonstrate that, given sufficient lag order, the approach accurately identifies the spatiotemporal origins of anomalies, while also quantifying the performance trade-offs introduced by mechanism approximation.

0 citationsRead paper

VCO-CARE: VCO-based Calibration-free Analog Readout for Electrodermal activity sensing

Sep 08, 2025

To address the challenges of high sensitivity, ultra-low power consumption, and calibration-free operation in skin electrodermal activity (EDA) sensing for wearable devices, this work proposes a calibration-free analog front-end (AFE) architecture based on a voltage-controlled oscillator (VCO). Instead of conventional resistance or current measurement paradigms requiring subject-specific calibration, the design employs frequency-domain readout to directly map conductance variations. It achieves 40 ps resolution and <0.0025% relative error over a 0–1.5 Hz bandwidth. Through low-noise circuit design and post-layout optimization, the system attains an average power consumption of only 2.3 μW and an input-referred noise of 0.8 μV_rms. The proposed AFE significantly mitigates inter-subject variability, enabling robust, ultra-low-power, on-chip EDA monitoring suitable for continuous, multi-user applications.

0 citationsRead paper

Effect Identification and Unit Categorization in the Multi-Score Regression Discontinuity Design with Application to LED Manufacturing

Aug 21, 2025

Real-world decision-making often relies on multi-dimensional threshold rules, yet conventional multi-score regression discontinuity designs (MRD) collapse these into a unidimensional framework, inducing noncompliance and biasing causal effect estimation. This paper establishes, for the first time, a classification framework for unit-level behavior under multi-dimensional cutoffs—explicitly defining compliers, avoiders, and inert agents. It introduces identification conditions for local average treatment effects at the sub-rule level, uncovering how rule decomposition shapes behavioral responses. We develop a novel MRD estimator leveraging multiple scores and validate it using both simulation studies and real-world semiconductor manufacturing data. Relative to standard MRD, our approach substantially reduces estimation variance and improves policy evaluation accuracy. Empirical application to LED production line optimization demonstrates its practical utility in industrial policy design.

0 citationsRead paper