Chien-Hung Lai
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Chien-Hung Lai has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
- Yuh-Shyan Hwang
Department of Electronic Engineering, National Taipei University of Technology, Taipei 10608, Taiwan
14 shared publications - Jiann-Jong Chen
Department of Electronic Engineering, National Taipei University of Technology, Taipei, Taiwan
9 shared publications - Yi Lin
Department of Business Administration, Takming University of Science and Technology, Taipei 11451, Taiwan
6 shared publications - Yitsen Ku
Department of Electrical and Computer Engineering, California State University at Fullerton, Fullerton, CA, USA
6 shared publications Joshua Ku
Department of Electrical Engineering, California State University, Fullerton, Fullerton, CA, USA
3 shared publications- Tzu-Shuang Liu
Department of Business Administration, Takming University of Science and Technology, Taipei 11451, Taiwan
2 shared publications
Latest Publications
- Open Access
- 30 September 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (19) , 9739
G-XCI: A Conditional Multi-View Audit of Stability and Cross-View Rank Agreement in Tabular Feature Attributions
Post hoc feature attributions are often interpreted from a single run, although their reproducibility and agreement with other explanatory views can depend on resampling, model family, representation, audit lens, output orientation, and inspection scope. We evaluate G-XCI as a conditional three-readout audit protocol. LAS measures reproducibility of an aggregate ranking derived from held-out Monte-Carlo Shapley attributions; GRS summarizes cross-seed variance of a specified global score profile; SCC measures Kendall tau-b agreement between the aggregate attribution ranking and a chosen global ranking. Across three tabular datasets, three model families, five stratified repeats, four global audit lenses, three inspection scopes, GRA ablations, and representation variants, Adult Income changed SCC sign with K, while audit-lens and representation changes materially altered SCC. Targeted supplementary audits showed that complementing the WDBC binary output reversed the mean GRA-based SCC sign in all nine model-scale combinations while leaving absolute local Shapley rankings unchanged, whereas forcing Standard GRA, absolute Spearman-to-output, and output permutation onto the same 128 held-out cases preserved all 27 original mean-SCC signs. A short lemma shows that robust-z followed by min-max equals direct min-max for nonconstant sequences. In the tested tabular binary-classification settings, G-XCI supports conditional auditing of aggregate-ranking reproducibility and cross-view rank agreement under a declared audit contract; it does not establish universal explanation correctness.
- 30 September 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (19) , 9739
Open AccessG-XCI: A Conditional Multi-View Audit of Stability and Cross-View Rank Agreement in Tabular Feature Attributions
Post hoc feature attributions are often interpreted from a single run, although their reproducibility and agreement with other explanatory views can depend on resampling, model family, representation, audit lens, output orientation, and inspection scope. We evaluate G-XCI as a conditional three-readout audit protocol. LAS measures reproducibility of an aggregate ranking derived from held-out Monte-Carlo Shapley attributions; GRS summarizes cross-seed variance of a specified global score profile; SCC measures Kendall tau-b agreement between the aggregate attribution ranking and a chosen global ranking. Across three tabular datasets, three model families, five stratified repeats, four global audit lenses, three inspection scopes, GRA ablations, and representation variants, Adult Income changed SCC sign with K, while audit-lens and representation changes materially altered SCC. Targeted supplementary audits showed that complementing the WDBC binary output reversed the mean GRA-based SCC sign in all nine model-scale combinations while leaving absolute local Shapley rankings unchanged, whereas forcing Standard GRA, absolute Spearman-to-output, and output permutation onto the same 128 held-out cases preserved all 27 original mean-SCC signs. A short lemma shows that robust-z followed by min-max equals direct min-max for nonconstant sequences. In the tested tabular binary-classification settings, G-XCI supports conditional auditing of aggregate-ranking reproducibility and cross-view rank agreement under a declared audit contract; it does not establish universal explanation correctness.
- Open Access
- 22 September 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (19) , 9418
Stability-Aware Auditing of Latent Regions in Sentence Embedding Spaces: Guarding Against Granularity Collapse in Unsupervised Text Analytics
Sentence-embedding clustering is often interpreted from a single partition, although a reproducible solution can still be excessively coarse, cardinality-variable, or representation-dependent. We present a stability-aware audit framework that separates label-free operating-point selection from post hoc taxonomy evaluation and explicitly distinguishes the original frozen confirmatory execution from separately frozen additional robustness analyses. On a pinned Banking77 test revision () and CLINC150+OOS (), the original anti-collapse rule selected seven-region MeanShift operating points with disjoint-random-stream confirmation ARIs of and , respectively. Additional paired refits showed that rejected coarse candidates can remain more stable than the selected solutions, and this stability–granularity inversion persisted under ARI, AMI, NMI, V-measure, and Fowlkes–Mallows similarity. The additional audits also showed that high membership similarity can coexist with refit-cardinality variation, that representation and broader pipeline perturbations materially affect the discovered structure, and that density-based stability must be interpreted jointly with coverage and cardinality. These results support a conservative conclusion: perturbation stability is evidence of reproducibility under a stated protocol, not proof of an appropriate resolution, semantic validity, or taxonomy recovery.- 22 September 2026
- Published by MDPI AG in Applied Sciences
- Vol. 16 (19) , 9418
Open AccessStability-Aware Auditing of Latent Regions in Sentence Embedding Spaces: Guarding Against Granularity Collapse in Unsupervised Text Analytics
Sentence-embedding clustering is often interpreted from a single partition, although a reproducible solution can still be excessively coarse, cardinality-variable, or representation-dependent. We present a stability-aware audit framework that separates label-free operating-point selection from post hoc taxonomy evaluation and explicitly distinguishes the original frozen confirmatory execution from separately frozen additional robustness analyses. On a pinned Banking77 test revision () and CLINC150+OOS (), the original anti-collapse rule selected seven-region MeanShift operating points with disjoint-random-stream confirmation ARIs of and , respectively. Additional paired refits showed that rejected coarse candidates can remain more stable than the selected solutions, and this stability–granularity inversion persisted under ARI, AMI, NMI, V-measure, and Fowlkes–Mallows similarity. The additional audits also showed that high membership similarity can coexist with refit-cardinality variation, that representation and broader pipeline perturbations materially affect the discovered structure, and that density-based stability must be interpreted jointly with coverage and cardinality. These results support a conservative conclusion: perturbation stability is evidence of reproducibility under a stated protocol, not proof of an appropriate resolution, semantic validity, or taxonomy recovery. - Open Access
- 17 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 47-54
CS-FXGBoost: Cost-Sensitive Focal Gradient Boosting for Rare-Event Classification under Extreme Class Imbalance
- 17 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 47-54
Open AccessCS-FXGBoost: Cost-Sensitive Focal Gradient Boosting for Rare-Event Classification under Extreme Class Imbalance
- Open Access
- 2 February 2026
- Published by MDPI AG in Engineering Proceedings
- Vol. 120 (1) , 29
Zero-Shot Complaint Classification and Style-Controlled Response Generation via Large Language Models for Emotion-Aware E-Commerce Review Management
We developed a large language model-powered system that classifies complaint categories and adapts response styles for e-commerce reviews. By integrating sentiment clustering, zero-shot classification, and style-conditioned prompt engineering, it enables context-aware, emotionally aligned reply generation for enhancing automated customer interaction and reputation management.
- 2 February 2026
- Published by MDPI AG in Engineering Proceedings
- Vol. 120 (1) , 29
Open AccessZero-Shot Complaint Classification and Style-Controlled Response Generation via Large Language Models for Emotion-Aware E-Commerce Review Management
We developed a large language model-powered system that classifies complaint categories and adapts response styles for e-commerce reviews. By integrating sentiment clustering, zero-shot classification, and style-conditioned prompt engineering, it enables context-aware, emotionally aligned reply generation for enhancing automated customer interaction and reputation management.
- Open Access
- 30 January 2026
- Published by MDPI AG in Engineering Proceedings
- Vol. 120 (1) , 13
Optimizing Discrete Wavelet Transform Watermarking with Genetic Algorithms for Resilient Digital Asset Protection Against Diverse Attacks
This study proposes a robust digital watermarking method combining the discrete wavelet transform and genetic algorithms (GAs) to enhance the protection of high-value digital assets against geometric attacks. By optimizing embedding strength through GA, this method achieves high imperceptibility and resilience under scaling, rotation, and translation. Experimental results demonstrate improved watermark recovery and visual fidelity, providing a practical solution for digital anti-counterfeiting.
- 30 January 2026
- Published by MDPI AG in Engineering Proceedings
- Vol. 120 (1) , 13
Open AccessOptimizing Discrete Wavelet Transform Watermarking with Genetic Algorithms for Resilient Digital Asset Protection Against Diverse Attacks
This study proposes a robust digital watermarking method combining the discrete wavelet transform and genetic algorithms (GAs) to enhance the protection of high-value digital assets against geometric attacks. By optimizing embedding strength through GA, this method achieves high imperceptibility and resilience under scaling, rotation, and translation. Experimental results demonstrate improved watermark recovery and visual fidelity, providing a practical solution for digital anti-counterfeiting.
- Open Access
- 25 December 2025
- Published by MDPI AG in Engineering Proceedings
- Vol. 120 (1) , 5
Composite Machine Learning System for Real-Time Response to Negative Online Reviews: A Case Study Based on the Negative Reinforcement Model of Digital Marketing
This research proposes a composite machine learning (ML) framework for real-time response to negative online reviews, grounded in the psychological principle of negative reinforcement. By integrating K-means clustering to group reviews by thematic similarity and bidirectional encoder representations from transformer (BERT)-based sentiment analysis to assess emotional tone, and the system identifies high-risk clusters requiring marketing intervention. Customized response strategies are designed based on cluster sentiment intensity, and their effectiveness can be evaluated via sentiment transformation functions. The proposed model provides a practical and adaptive approach to digital marketing, enabling brands to respond rapidly, reduce dissatisfaction, and enhance consumer trust in a data-driven environment.
- 25 December 2025
- Published by MDPI AG in Engineering Proceedings
- Vol. 120 (1) , 5
Open AccessComposite Machine Learning System for Real-Time Response to Negative Online Reviews: A Case Study Based on the Negative Reinforcement Model of Digital Marketing
This research proposes a composite machine learning (ML) framework for real-time response to negative online reviews, grounded in the psychological principle of negative reinforcement. By integrating K-means clustering to group reviews by thematic similarity and bidirectional encoder representations from transformer (BERT)-based sentiment analysis to assess emotional tone, and the system identifies high-risk clusters requiring marketing intervention. Customized response strategies are designed based on cluster sentiment intensity, and their effectiveness can be evaluated via sentiment transformation functions. The proposed model provides a practical and adaptive approach to digital marketing, enabling brands to respond rapidly, reduce dissatisfaction, and enhance consumer trust in a data-driven environment.
- Open Access
- 3 March 2025
- Published by MDPI AG in Engineering Proceedings
- Vol. 89 (1) , 32
The Use of Support Vector Machine to Classify Potential Customers for the Wealth Management of a Bank
We developed a method for the evaluation and selection of customer business analysis in two stages. First, using the bank’s existing expert model, artificial rules of thumb were used to evaluate the value of each field of the data and establish screening rules. Secondly, the machine learning feature screening method was applied based on the customer’s transaction data to find out whether the customer’s contribution to the bank had a significant impact as a feature of the model. Based on the results, the best classification model was selected through data verification. The effectiveness of the proposed model was validated through actual case analysis, taking wealth management in banks as an example. The classification method, using support vector machines (SVMs), effectively assists banks in identifying potential customers efficiently and in planning to manage customers. This method helps to avoid the traditional blind spots, which emerge based on subjective judgment, and allows bank wealth managers to promote customer relationship management (CRM).
- 3 March 2025
- Published by MDPI AG in Engineering Proceedings
- Vol. 89 (1) , 32
Open AccessThe Use of Support Vector Machine to Classify Potential Customers for the Wealth Management of a Bank
We developed a method for the evaluation and selection of customer business analysis in two stages. First, using the bank’s existing expert model, artificial rules of thumb were used to evaluate the value of each field of the data and establish screening rules. Secondly, the machine learning feature screening method was applied based on the customer’s transaction data to find out whether the customer’s contribution to the bank had a significant impact as a feature of the model. Based on the results, the best classification model was selected through data verification. The effectiveness of the proposed model was validated through actual case analysis, taking wealth management in banks as an example. The classification method, using support vector machines (SVMs), effectively assists banks in identifying potential customers efficiently and in planning to manage customers. This method helps to avoid the traditional blind spots, which emerge based on subjective judgment, and allows bank wealth managers to promote customer relationship management (CRM).
- Open Access
- 24 April 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 41644-41653
A New Low-Noise I-Squared Buck Converter Suitable for Wireless Sensor Networks With Dual-Current-Acceleration Techniques
This paper proposed a new low-noise I-squard 2 nd -order continuous-time-delta-sigma-modulation (CT-DSM) buck converter suitable for wireless sensor networks with dual-current-acceleration (DCA) techniques. First, it adopts the second-order CT-DSM scheme, which can effectively utilize the oversampling and the noise shaping techniques to disperse the noise and push it to higher frequencies. Second, the use of the pseudo-inductor-current-sensor can increase the bandwidth of the closed-loop gain and consequently speeds up the transient response of the converter. Third, the dynamic-current regulator utilizes the principle of current charging and discharging to make the converter respond faster when the load changes dynamically. Fourth, this paper combines the pseudo-inductor-current-sensor and the dynamic-current regulator to realize the dual-current-acceleration technique, which can accelerate the transient response by up to 45.5%. The proposed converter is designed and fabricated in TSMC 0.18μm 1P6M CMOS process, and the chip area is 1.065 mm 2 . When the load current changes from 50mA to 500mA and 500mA to 50mA, the transient responses are 2.4μs and 3μs, respectively, and the peak efficiency is 92.4%. The measurement result of ONR is 78.087dB, with the sampling frequency of 10 MHz.
- 24 April 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 41644-41653
Open AccessA New Low-Noise I-Squared Buck Converter Suitable for Wireless Sensor Networks With Dual-Current-Acceleration Techniques
This paper proposed a new low-noise I-squard 2 nd -order continuous-time-delta-sigma-modulation (CT-DSM) buck converter suitable for wireless sensor networks with dual-current-acceleration (DCA) techniques. First, it adopts the second-order CT-DSM scheme, which can effectively utilize the oversampling and the noise shaping techniques to disperse the noise and push it to higher frequencies. Second, the use of the pseudo-inductor-current-sensor can increase the bandwidth of the closed-loop gain and consequently speeds up the transient response of the converter. Third, the dynamic-current regulator utilizes the principle of current charging and discharging to make the converter respond faster when the load changes dynamically. Fourth, this paper combines the pseudo-inductor-current-sensor and the dynamic-current regulator to realize the dual-current-acceleration technique, which can accelerate the transient response by up to 45.5%. The proposed converter is designed and fabricated in TSMC 0.18μm 1P6M CMOS process, and the chip area is 1.065 mm 2 . When the load current changes from 50mA to 500mA and 500mA to 50mA, the transient responses are 2.4μs and 3μs, respectively, and the peak efficiency is 92.4%. The measurement result of ONR is 78.087dB, with the sampling frequency of 10 MHz.
- 17 August 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 22 (18) , 18202-18209
A New Ultra-Fast-Response Low-Transient-Voltage Boost Converter Suitable for Low-Voltage Solar Cells in Wireless Sensor Networks
In this article, a new ultra-fast-response low-transient-voltage boost converter suitable for low-voltage solar cells in wireless sensor networks is presented. The proposed converter not only achieves a faster transient response but also higher power conversion efficiency. With the current-feedback adaptive-controlled technique, the system can operate in the more fixed switching frequency without extra phase-locked-loop circuits. The proposed new rail-to-rail current sensing circuit, which produces the inductor current information to the controlled loop, not only increases the stability of the system but also improves the dynamic response. The addition of a zero-current detector (ZCD) prevents the reverse current as the converter operates at light load, which enhances the conversion efficiency, and ensures that the converter works stably in the discontinuous conduction mode (DCM). The proposed converter has been implemented with TSMC 0.18- $ {\mu }\text{m}$ 1P6M CMOS processes and occupies an area of 1.19 mm $\times1.2$ mm. The measured results show that the transient response is $2 ~{\mu }\text{s}$ when the load current changes between 1 and 200 mA. The peak conversion efficiency is 92.3% under 140-mA load current. The maximum efficiency improvement is 14.8% at 5-mA load current. The major contributions of this article are listed in the following: 1) we proposed a new architecture of boost converters with improved current-feedback and adaptive-controlled techniques; 2) the proposed architecture has one current loop and one voltage loop to balance the output voltage and the output current; 3) the proposed boost converter has the characteristics of fast response time and low transient voltage; 4) we have designed, implemented, and measured the proposed converters; and 5) from the figure of merit (FOM) formula, the performance of the proposed boost converter is better than other recent works.
- 17 August 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 22 (18) , 18202-18209
A New Ultra-Fast-Response Low-Transient-Voltage Boost Converter Suitable for Low-Voltage Solar Cells in Wireless Sensor Networks
In this article, a new ultra-fast-response low-transient-voltage boost converter suitable for low-voltage solar cells in wireless sensor networks is presented. The proposed converter not only achieves a faster transient response but also higher power conversion efficiency. With the current-feedback adaptive-controlled technique, the system can operate in the more fixed switching frequency without extra phase-locked-loop circuits. The proposed new rail-to-rail current sensing circuit, which produces the inductor current information to the controlled loop, not only increases the stability of the system but also improves the dynamic response. The addition of a zero-current detector (ZCD) prevents the reverse current as the converter operates at light load, which enhances the conversion efficiency, and ensures that the converter works stably in the discontinuous conduction mode (DCM). The proposed converter has been implemented with TSMC 0.18- $ {\mu }\text{m}$ 1P6M CMOS processes and occupies an area of 1.19 mm $\times1.2$ mm. The measured results show that the transient response is $2 ~{\mu }\text{s}$ when the load current changes between 1 and 200 mA. The peak conversion efficiency is 92.3% under 140-mA load current. The maximum efficiency improvement is 14.8% at 5-mA load current. The major contributions of this article are listed in the following: 1) we proposed a new architecture of boost converters with improved current-feedback and adaptive-controlled techniques; 2) the proposed architecture has one current loop and one voltage loop to balance the output voltage and the output current; 3) the proposed boost converter has the characteristics of fast response time and low transient voltage; 4) we have designed, implemented, and measured the proposed converters; and 5) from the figure of merit (FOM) formula, the performance of the proposed boost converter is better than other recent works.
- Open Access
- 14 June 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 10, 63063-63072
A Low-Noise Fast-Transient-Response Delta-Sigma-Modulation Buck Converter With Hysteresis-Voltage-Controlled Techniques
A low-noise fast-transient-response second-order delta-sigma-modulation buck converter with hysteresis-voltage-controlled techniques is proposed. With the proposed control approach, the transient time can be accelerated by roughly 60%. The rail-to-rail OTA generates a current ISense, which replicates the inductor current IL with K times. As ISense flows through the capacitor CSense, it will be converted into VSense. Then, the hysteresis-voltage-controlled (HVC) circuit compares the two terminal voltages of hysteresis comparator to detect the overshoot and undershoot of VSense. Once VSense is detected, the output signal of HVC circuits becomes opposite to the previous state to conduct MP and MN previously. Besides, the 2nd-order delta-sigma-modulation (DSM) circuit plays a vital role of mitigating noise-interference and elevating noise in whole circuits. The proposed converter has been fabricated in TSMC 0.18μm 1P6M CMOS processes with an active area of 1.19×1.09mm2. The measured results show the transient time are 3.5μs and 3.2μs, respectively, when the load current changes between 500mA and 100mA. On the basic of the measured results of fast-fourier-transform (FFT), the value of output-to-noise ratio (ONR) is 76.6dB at the sampling frequency of 10MHz. The peak conversion efficiency is 92.1% while the load current is 300mA.
- 14 June 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 10, 63063-63072
Open AccessA Low-Noise Fast-Transient-Response Delta-Sigma-Modulation Buck Converter With Hysteresis-Voltage-Controlled Techniques
A low-noise fast-transient-response second-order delta-sigma-modulation buck converter with hysteresis-voltage-controlled techniques is proposed. With the proposed control approach, the transient time can be accelerated by roughly 60%. The rail-to-rail OTA generates a current ISense, which replicates the inductor current IL with K times. As ISense flows through the capacitor CSense, it will be converted into VSense. Then, the hysteresis-voltage-controlled (HVC) circuit compares the two terminal voltages of hysteresis comparator to detect the overshoot and undershoot of VSense. Once VSense is detected, the output signal of HVC circuits becomes opposite to the previous state to conduct MP and MN previously. Besides, the 2nd-order delta-sigma-modulation (DSM) circuit plays a vital role of mitigating noise-interference and elevating noise in whole circuits. The proposed converter has been fabricated in TSMC 0.18μm 1P6M CMOS processes with an active area of 1.19×1.09mm2. The measured results show the transient time are 3.5μs and 3.2μs, respectively, when the load current changes between 500mA and 100mA. On the basic of the measured results of fast-fourier-transform (FFT), the value of output-to-noise ratio (ONR) is 76.6dB at the sampling frequency of 10MHz. The peak conversion efficiency is 92.1% while the load current is 300mA.