Wen-Hung Liao
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Wen-Hung Liao has a total of 212 co-authors. Here are the top co-authors by number of shared publications:
J.K. Aggarwal
Dept. of Electr. & Comput. Eng., Texas Univ., Austin, TX, USA
4 shared publications- Yi-Chieh Wu
National Chengchi University
4 shared publications Po-Ming Chen
National Chengchi University
3 shared publicationsTsai-Yen Li
3 shared publicationsB. Sabata
Dept. of Electr. & Comput. Eng., Texas Univ., Austin, TX, USA
2 shared publications- Chien-Ming Yang
Department of Psychology National Cheng‐Chi University Taipei City Taiwan
2 shared publications
Latest Publications
- 3 August 2026
- Published by Springer Nature
- p. 48-62
Decoding-Time Fusion of OCR and Large Language Models for Traditional Chinese Historical Document Recognition
- 3 August 2026
- Published by Springer Nature
- p. 48-62
Decoding-Time Fusion of OCR and Large Language Models for Traditional Chinese Historical Document Recognition
- Open Access
On the state of protein function prediction: a report on the fourth CAFA challenge
Background: The Critical Assessment of Functional Annotation (CAFA) is a community effort held to understand the field of computational protein function prediction. Every three years, since 2010, the organizers initiate an experiment to collect function predictions on a large set of proteins and then evaluate the performance of predicting methods on a subset of proteins that have accumulated experimental annotations between the submission deadline and the evaluation time. CAFA provides an independent and rigorous assessment of the current state of the art, thus leveling the playing field, highlighting successes, revealing bottlenecks, and offering a forum for the exchange of ideas in protein science. Here, we report the results of the fourth CAFA experiment (CAFA4). Results: CAFA4 featured the participation of 148 methods from 70 research groups on a total of 46,205 unique proteins over a 5-year annotation accumulation phase, the longest in any CAFA. In a comparison across CAFA2-CAFA4 methods, the prediction of Gene Ontology (GO) terms has clearly improved across all three GO aspects and traditional evaluation settings. While not achieving the first rank, several CAFA2 and CAFA3 methods featured in the top ten methods in many evaluations, suggesting that earlier methods still hold relevance. The performance is weaker in the newly introduced “partial knowledge” evaluation category (proteins with experimental annotations before submission deadline that gained additional annotations in the same GO aspect during the annotation accumulation phase), highlighting the need for a new class of methods. The rankings of the methods were stable over the years in traditional evaluation settings, but less so in the new partial knowledge evaluation. Overall, the field continues to progress with some influx of new participants. Sustained efforts will be necessary to substantially advance it.
Open AccessOn the state of protein function prediction: a report on the fourth CAFA challenge
Background: The Critical Assessment of Functional Annotation (CAFA) is a community effort held to understand the field of computational protein function prediction. Every three years, since 2010, the organizers initiate an experiment to collect function predictions on a large set of proteins and then evaluate the performance of predicting methods on a subset of proteins that have accumulated experimental annotations between the submission deadline and the evaluation time. CAFA provides an independent and rigorous assessment of the current state of the art, thus leveling the playing field, highlighting successes, revealing bottlenecks, and offering a forum for the exchange of ideas in protein science. Here, we report the results of the fourth CAFA experiment (CAFA4). Results: CAFA4 featured the participation of 148 methods from 70 research groups on a total of 46,205 unique proteins over a 5-year annotation accumulation phase, the longest in any CAFA. In a comparison across CAFA2-CAFA4 methods, the prediction of Gene Ontology (GO) terms has clearly improved across all three GO aspects and traditional evaluation settings. While not achieving the first rank, several CAFA2 and CAFA3 methods featured in the top ten methods in many evaluations, suggesting that earlier methods still hold relevance. The performance is weaker in the newly introduced “partial knowledge” evaluation category (proteins with experimental annotations before submission deadline that gained additional annotations in the same GO aspect during the annotation accumulation phase), highlighting the need for a new class of methods. The rankings of the methods were stable over the years in traditional evaluation settings, but less so in the new partial knowledge evaluation. Overall, the field continues to progress with some influx of new participants. Sustained efforts will be necessary to substantially advance it.
- 9 November 2025
- Published by Springer Nature
- p. 265-280
Deep Learning-Based Restoration of Voice-Converted Audio for Speech and Speaker Recognition
- 9 November 2025
- Published by Springer Nature
- p. 265-280
Deep Learning-Based Restoration of Voice-Converted Audio for Speech and Speaker Recognition
- 9 November 2025
- Published by Springer Nature
- p. 250-264
Selective Freezing of Feature Hierarchies in Deep Models for Machine Unlearning
- 9 November 2025
- Published by Springer Nature
- p. 250-264
Selective Freezing of Feature Hierarchies in Deep Models for Machine Unlearning
- 28 January 2024
- Published by Springer Nature
- p. 503-516
Adversarially Robust Deepfake Detection via Adversarial Feature Similarity Learning
- 28 January 2024
- Published by Springer Nature
- p. 503-516
Adversarially Robust Deepfake Detection via Adversarial Feature Similarity Learning
- Open Access
Towards Adversarial Robustness for Multi-Mode Data through Metric Learning
Adversarial attacks have become one of the most serious security issues in widely used deep neural networks. Even though real-world datasets usually have large intra-variations or multiple modes, most adversarial defense methods, such as adversarial training, which is currently one of the most effective defense methods, mainly focus on the single-mode setting and thus fail to capture the full data representation to defend against adversarial attacks. To confront this challenge, we propose a novel multi-prototype metric learning regularization for adversarial training which can effectively enhance the defense capability of adversarial training by preventing the latent representation of the adversarial example changing a lot from its clean one. With extensive experiments on CIFAR10, CIFAR100, MNIST, and Tiny ImageNet, the evaluation results show the proposed method improves the performance of different state-of-the-art adversarial training methods without additional computational cost. Furthermore, besides Tiny ImageNet, in the multi-prototype CIFAR10 and CIFAR100 where we reorganize the whole datasets of CIFAR10 and CIFAR100 into two and ten classes, respectively, the proposed method outperforms the state-of-the-art approach by 2.22% and 1.65%, respectively. Furthermore, the proposed multi-prototype method also outperforms its single-prototype version and other commonly used deep metric learning approaches as regularization for adversarial training and thus further demonstrates its effectiveness.
Open AccessTowards Adversarial Robustness for Multi-Mode Data through Metric Learning
Adversarial attacks have become one of the most serious security issues in widely used deep neural networks. Even though real-world datasets usually have large intra-variations or multiple modes, most adversarial defense methods, such as adversarial training, which is currently one of the most effective defense methods, mainly focus on the single-mode setting and thus fail to capture the full data representation to defend against adversarial attacks. To confront this challenge, we propose a novel multi-prototype metric learning regularization for adversarial training which can effectively enhance the defense capability of adversarial training by preventing the latent representation of the adversarial example changing a lot from its clean one. With extensive experiments on CIFAR10, CIFAR100, MNIST, and Tiny ImageNet, the evaluation results show the proposed method improves the performance of different state-of-the-art adversarial training methods without additional computational cost. Furthermore, besides Tiny ImageNet, in the multi-prototype CIFAR10 and CIFAR100 where we reorganize the whole datasets of CIFAR10 and CIFAR100 into two and ten classes, respectively, the proposed method outperforms the state-of-the-art approach by 2.22% and 1.65%, respectively. Furthermore, the proposed multi-prototype method also outperforms its single-prototype version and other commonly used deep metric learning approaches as regularization for adversarial training and thus further demonstrates its effectiveness.
- 3 January 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Detection of Synthesized Satellite Images Using Deep Neural Networks
The technology of generative adversarial networks (GAN) is constantly evolving, and synthesized images can no longer be accurately distinguished by the human eyes alone. GAN has been applied to the analysis of satellite images, mostly for the purpose of data augmentation. Recently, however, we have seen a twist in its usage. In information warfare, GAN has been used to create fake satellite images or modify the image content by putting fake bridges, buildings and clouds to mislead or conceal important intelligence. To address the increasing counterfeit cases in satellite images, the goal of this research is to develop algorithms that can classify fake remote sensing images robustly and efficiently. There exist many techniques to synthesize or manipulate the content of satellite images. In this paper, we focus on the case when the entire image is forged. Three satellite image synthesis methods, including ProGAN, cGAN and CycleGAN will be investigated. The effect of image pre-processing such as histogram equalization and bilateral filter will also be evaluated. Experiments show that satellite images generated by different GANs can be easily identified by individually trained models. The performance degraded when model trained with one type of GAN samples is employed to determine the originality of images synthesized with other types of GANs. Additionally, when histogram equalization is applied to the images, the detection model fails to distinguish its authenticity. A four-class universal classification model is proposed to address this issue. An overall accuracy of over 99% has been achieved even when pre-processing has been applied.
- 3 January 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
Detection of Synthesized Satellite Images Using Deep Neural Networks
The technology of generative adversarial networks (GAN) is constantly evolving, and synthesized images can no longer be accurately distinguished by the human eyes alone. GAN has been applied to the analysis of satellite images, mostly for the purpose of data augmentation. Recently, however, we have seen a twist in its usage. In information warfare, GAN has been used to create fake satellite images or modify the image content by putting fake bridges, buildings and clouds to mislead or conceal important intelligence. To address the increasing counterfeit cases in satellite images, the goal of this research is to develop algorithms that can classify fake remote sensing images robustly and efficiently. There exist many techniques to synthesize or manipulate the content of satellite images. In this paper, we focus on the case when the entire image is forged. Three satellite image synthesis methods, including ProGAN, cGAN and CycleGAN will be investigated. The effect of image pre-processing such as histogram equalization and bilateral filter will also be evaluated. Experiments show that satellite images generated by different GANs can be easily identified by individually trained models. The performance degraded when model trained with one type of GAN samples is employed to determine the originality of images synthesized with other types of GANs. Additionally, when histogram equalization is applied to the images, the detection model fails to distinguish its authenticity. A four-class universal classification model is proposed to address this issue. An overall accuracy of over 99% has been achieved even when pre-processing has been applied.
- 3 July 2021
- Published by Springer Nature
- p. 377-393
Investigating Viewer’s Reliance on Captions Based on Gaze Information
- 3 July 2021
- Published by Springer Nature
- p. 377-393
Investigating Viewer’s Reliance on Captions Based on Gaze Information
- Open Access
- 18 November 2020
- Published by Springer Nature in BMC Bioinformatics
- Vol. 21 (S6) , 1-16
GODoc: high-throughput protein function prediction using novel k-nearest-neighbor and voting algorithms
Background: Biological data has grown explosively with the advance of next-generation sequencing. However, annotating protein function with wet lab experiments is time-consuming. Fortunately, computational function prediction can help wet labs formulate biological hypotheses and prioritize experiments. Gene Ontology (GO) is a framework for unifying the representation of protein function in a hierarchical tree composed of GO terms. Results: We propose GODoc, a general protein GO prediction framework based on sequence information which combines feature engineering, feature reduction, and a novel k-nearest-neighbor algorithm to resolve the multiple GO prediction problem. Comprehensive evaluation on CAFA2 shows that GODoc performs better than two baseline models. In the CAFA3 competition (68 teams), GODoc ranks 10th in Cellular Component Ontology. Regarding the species-specific task, the proposed method ranks 10th and 8th in the eukaryotic Cellular Component Ontology and the prokaryotic Molecular Function Ontology, respectively. In the term-centric task, GODoc performs third and is tied for first for the biofilm formation of Pseudomonas aeruginosa and the long-term memory of Drosophila melanogaster, respectively. Conclusions: We have developed a novel and effective strategy to incorporate a training procedure into the k-nearest neighbor algorithm (instance-based learning) which is capable of solving the Gene Ontology multiple-label prediction problem, which is especially notable given the thousands of Gene Ontology terms.
- 18 November 2020
- Published by Springer Nature in BMC Bioinformatics
- Vol. 21 (S6) , 1-16
Open AccessGODoc: high-throughput protein function prediction using novel k-nearest-neighbor and voting algorithms
Background: Biological data has grown explosively with the advance of next-generation sequencing. However, annotating protein function with wet lab experiments is time-consuming. Fortunately, computational function prediction can help wet labs formulate biological hypotheses and prioritize experiments. Gene Ontology (GO) is a framework for unifying the representation of protein function in a hierarchical tree composed of GO terms. Results: We propose GODoc, a general protein GO prediction framework based on sequence information which combines feature engineering, feature reduction, and a novel k-nearest-neighbor algorithm to resolve the multiple GO prediction problem. Comprehensive evaluation on CAFA2 shows that GODoc performs better than two baseline models. In the CAFA3 competition (68 teams), GODoc ranks 10th in Cellular Component Ontology. Regarding the species-specific task, the proposed method ranks 10th and 8th in the eukaryotic Cellular Component Ontology and the prokaryotic Molecular Function Ontology, respectively. In the term-centric task, GODoc performs third and is tied for first for the biofilm formation of Pseudomonas aeruginosa and the long-term memory of Drosophila melanogaster, respectively. Conclusions: We have developed a novel and effective strategy to incorporate a training procedure into the k-nearest neighbor algorithm (instance-based learning) which is capable of solving the Gene Ontology multiple-label prediction problem, which is especially notable given the thousands of Gene Ontology terms.
- Open Access
- 9 October 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 8, 186976-186992
Analysis of Learning Patterns and Performance—A Case Study of 3-D Modeling Lessons in the K-12 Classrooms
As more schools incorporate 3D printing into their curriculum to stimulate the creativity of K-12 students with a learning-by-doing approach, it becomes crucial to understand how users work with 3D modeling tools and to evaluate integrated lesson plans in the STEAM (Science, Technology, Engineering, Arts and Math) educational framework. Our work consists of two stages: an investigation of usage patterns of modeling, and an evaluation of the usability of Qmodel Creator, in collaboration with Lanyu Primary and Junior High School, Sanchong High School, and Affiliated Experimental Elementary School of National Chengchi University. Participants operation logs, screen recordings, and finished work for respective 3D modeling software were recorded and analyzed. Two types of indicators have been developed. One is concerned with the quantification of learning behavior, including Effective Operating Period (EOP), Trial-and-Error Period (TEP), and Implementation Period (IP). The other has to do with the evaluation of learning outcome, i.e., the complexity of 3D models, including the Degree of Detail (DoD), shape (Cf), partition (Cp), and block-ratio (Cr) complexity. Based on the proposed features, we are able to identify the key factors that affect students’ learning experiences and performance in terms of learning patterns and model completeness. Through these indicators, instructors can gain better insights into student’s learning status of 3D modeling software.
- 9 October 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 8, 186976-186992
Open AccessAnalysis of Learning Patterns and Performance—A Case Study of 3-D Modeling Lessons in the K-12 Classrooms
As more schools incorporate 3D printing into their curriculum to stimulate the creativity of K-12 students with a learning-by-doing approach, it becomes crucial to understand how users work with 3D modeling tools and to evaluate integrated lesson plans in the STEAM (Science, Technology, Engineering, Arts and Math) educational framework. Our work consists of two stages: an investigation of usage patterns of modeling, and an evaluation of the usability of Qmodel Creator, in collaboration with Lanyu Primary and Junior High School, Sanchong High School, and Affiliated Experimental Elementary School of National Chengchi University. Participants operation logs, screen recordings, and finished work for respective 3D modeling software were recorded and analyzed. Two types of indicators have been developed. One is concerned with the quantification of learning behavior, including Effective Operating Period (EOP), Trial-and-Error Period (TEP), and Implementation Period (IP). The other has to do with the evaluation of learning outcome, i.e., the complexity of 3D models, including the Degree of Detail (DoD), shape (Cf), partition (Cp), and block-ratio (Cr) complexity. Based on the proposed features, we are able to identify the key factors that affect students’ learning experiences and performance in terms of learning patterns and model completeness. Through these indicators, instructors can gain better insights into student’s learning status of 3D modeling software.