Rua-huan Tsaih | National chenchi University (original) (raw)

Papers by Rua-huan Tsaih

Research paper thumbnail of The AI Tech-Stack Model

The AI Tech-Stack Model

Communications of the ACM

Management and technology challenges of AI-enabled application projects.

Research paper thumbnail of The Cramming, Softening and Integrating Learning Algorithm with Parametric ReLU Activation Function for Binary Input/Output Problems

The Cramming, Softening and Integrating Learning Algorithm with Parametric ReLU Activation Function for Binary Input/Output Problems

2019 International Joint Conference on Neural Networks (IJCNN), 2019

Rare Artificial Neural Networks studies address simultaneously the challenges of (1) systematical... more Rare Artificial Neural Networks studies address simultaneously the challenges of (1) systematically adjusting the amount of used hidden layer nodes within the learning process, (2) adopting Parametric ReLU activation function instead of tanh function for fast learning, and (3) guaranteeing learning all training data. This study will address these challenges through deriving the CSI (Cramming, Softening and Integrating) learning algorithm for the single-hidden layer feed-forward neural networks with the binary input/output and making the technical justification. To further verify the proposed learning algorithm, this study conducts an empirical experiment using SPECT heart diagnosis data set from UCI Machine Learning repository. The learning algorithm is implemented via the advanced TensorFlow and GPU.

Research paper thumbnail of The ICT predicament of new ICT-enabled service

ArXiv, 2015

The advancement of information and communication technologies (ICT) has triggered many ICT-enable... more The advancement of information and communication technologies (ICT) has triggered many ICT-enabled services. Regarding this service, the complementary ICT system involves with customers' devices, industry-wide ICT development and nation-wide ICT infrastructure, which are difficult for any individual organization to control. The ICT predicament is the phenomenon that the complementary ICT system is inferior in delivering the promised service quality of new ICT-enabled service. With the ICT predicament, companies face the decision-making dilemma in launching the new service or postponing the launch. This study proposes a process to resolve the decision-making dilemma regarding the ICT predicament.

Research paper thumbnail of Using Virtual Reality for Museum Exhibitions: The Effects of Attention and Engagement for National Palace Museum

Using Virtual Reality for Museum Exhibitions: The Effects of Attention and Engagement for National Palace Museum

Information technologies provide important opportunities for museums to create more engaging visi... more Information technologies provide important opportunities for museums to create more engaging visitor experience. This study collaborates with National Palace Museum and focuses on investigating whether virtual reality is an effective communication medium for museum exhibition. We compare three communication mediums-- video, website and virtual reality and investigate their impacts on user experience, in terms of engagement and attention. We are also interested in the role of users’ personality and their prior experience of communication media in moderating these effects. The results of findings show that virtual reality can catch more attention from user, and attention enforces their engagement, which leads to better user experience. Furthermore, personality and relative experience are both insignificant in mediating the relationship between engagement and visit intention, but we found that when using communication medium, extravert can have higher attention than the other users, which deserve more study in the future

Research paper thumbnail of How Design Features Lead to Visitors' Visit Intention through Virtual Reality Experience: The Case of National Palace Museum

How Design Features Lead to Visitors' Visit Intention through Virtual Reality Experience: The Case of National Palace Museum

The growing number of museums attempts to apply virtual reality technology to enhance the visitor... more The growing number of museums attempts to apply virtual reality technology to enhance the visitors’ experience, and significant research effort has been also made in the area of modeling the relics in VR. However, the studies on the real effects of the VR exhibition are less. In this paper, we draw upon the stimulus–organism–response (S- O-R) framework to theorize how system design features stimuli make an effect on the visitors’ experience of National Palace Museum’s VR exhibition, which in turn impacts the visit intention. Three types of design features are examined: interactivity, vividness, and realism. Additionally, we are interested in two types of VR experience: immersion and involvement. Furthermore, the study compares the VR effects across different kinds of relics (artifact, painting and calligraphy). The results of findings show that all of design features impact on immersion level for all relic types. Only for the calligraphy, the vividness impact on the involvement level, and only for the artifact, the interactivity impact on the involvement level

Research paper thumbnail of The Mathematical Programming and the Rule Ex-traction from Layered Feed-forward Neural Net-works

The mathematical programming analysis, instead of a data analysis, is proposed for identifying th... more The mathematical programming analysis, instead of a data analysis, is proposed for identifying the convex polyhedron associated with each rule. The area depicted in the rule premise covers a convex polyhedron in the input space, and the adopted approximation function for the output value is a multivariate polynomial function of x, the outside stimulus input. Moreover, the mathematical programming analysis is proposed for examining the extracted rules to explore features.

Research paper thumbnail of An improved back propagation neural network learning algorithm

An improved back propagation neural network learning algorithm

Artificial neural networks are dynamic computing systems which are made up of a number of simple,... more Artificial neural networks are dynamic computing systems which are made up of a number of simple, highly interconnected processing elements. They are designed to imitate some aspects of brain function such that they can self-organize, learn, and react. With these characteristics, artificial neural networks resemble those in the brain more closely than had previously been constructed. Yet the slowness of the convergence speed of learning (C.S.L.) undermines their applications to practical problems. Learning governed by an iterative optimization algorithm is a matter of searching for proper strengths of connections (weights) such that, at the end of learning, the network responds correctly to all training stimuli. For example, the Back Propagation learning algorithm (the BP algorithm) uses the gradient descent procedure to search for proper weights that minimize an objective function which is a measurement of the discrepancy between actual outputs and desired outputs with respect to all training stimuli. The research objective is to accelerate the C.S.L. We have studied two layer feed-forward artificial neural networks with one output node. We have achieved three things: extensive experimental observations of various 2-classes categorization problems with learning using the BP algorithm, explanations of these observations, and acceleration of the convergence speed of learning. Experimental observations show that there are some "attractors" in the landscape of the objective function that often interfere with the search process and cause the slowness of the C.S.L. These attractors are related to the existence of saddle stationary points. An improved version of the learning algorithm has been developed based on these observations. Experimental results obtained on a class of problems called parity problems have shown that this improved version is at least 15 times faster than using the conventional method which combined the momentum version of the gradient descent method and the adaptive stepsize technique.

Research paper thumbnail of The Research of Multi-Layer Topic Map Analysis using Co-word Analysis with Growing Hierarchical Self-organizing Map

International Journal of Digital Content Technology and its Applications, 2011

The purpose of this study was to propose a multi-layer topic map analysis using co-word analysis ... more The purpose of this study was to propose a multi-layer topic map analysis using co-word analysis of informetrics with Growing Hierarchical Self-Organizing Map (GHSOM). The topic map illustrated the delicate intertwining of subject areas and provided a more explicit illustration of the concepts within each subject area. We applied GHSOM, a text-mining Neural Networks tool, to obtain a hierarchical topic map. After taking up one example of altruism in evaluation, we suggest that topic map may disclose some important facts from a whole bunch of data.

Research paper thumbnail of A resistant learning procedure for coping with outliers

A resistant learning procedure for coping with outliers

Annals of Mathematics and Artificial Intelligence, 2009

In the context of resistant learning, outliers are the observations far away from the fitting fun... more In the context of resistant learning, outliers are the observations far away from the fitting function that is deduced from a subset of the given observations and whose form is adaptable during the process. This study presents a resistant learning procedure for coping with outliers via single-hidden layer feed-forward neural network (SLFN). The smallest trimmed sum of squared residuals principle

Research paper thumbnail of Artificial Intelligence in Smart Tourism: A Conceptual Framework

Smart tourism destination as: an innovative tourist destination, built on an infrastructure of st... more Smart tourism destination as: an innovative tourist destination, built on an infrastructure of state-of-the-art technology guaranteeing the sustainable development of tourist areas, accessible to everyone, which facilitates the visitor’s interaction with and integration into his or her surroundings, increases the quality of the experience at the destination, and improves residents’ quality of life. Lopez de Avila (2015). Smart tourism involves multiple components and layers of “smart” include (1) Smart Destinations which was special cases of smart cities integration of ICT’s into physical infrastructure, (2) Smart experience which specifically focus on technology-mediated tourism experience and their engagement through personalization, context-awareness and real-time monitoring, (3) Smart business refer to the complex business ecosystem that creates and supports the exchange of touristic resource and the co-creation of tourism experience. Gretzel et al, (2015). Smart tourism also cl...

Research paper thumbnail of The New Method of Feature Selection for Intradialytic Hypotension Prediction using Machine learning

The New Method of Feature Selection for Intradialytic Hypotension Prediction using Machine learning

2021 IEEE 3rd Eurasia Conference on Biomedical Engineering, Healthcare and Sustainability (ECBIOS)

Intradialytic hypotension (IDH) is a common complication and make patient get brain and heart dis... more Intradialytic hypotension (IDH) is a common complication and make patient get brain and heart disease. This study aimed to use the new method of feature selection to improve the performance of IDH prediction. We used feature importance by a different model to sort from high to low one by one, remove the lowest feature importance, and follow the rule again. Therefore, we obtain the combination of feature values that have the highest accuracy for predictive power. After applying REFCV with linear regression, the AUC of ROC increased to 0.966 (2-layers LSTM) and 0.965 (3-layers LSTM). The feature selection method we designed resulted in a better prediction ability in the model.

Research paper thumbnail of ANN Mechanism for Network Traffic Anomaly Detection in the Concept Drifting Environment

ANN Mechanism for Network Traffic Anomaly Detection in the Concept Drifting Environment

2018 IEEE Conference on Dependable and Secure Computing (DSC)

Research paper thumbnail of Managing Innovation and Cultural Management in the Digital Era

Managing Innovation and Cultural Management in the Digital Era

Research paper thumbnail of Quantitative quality estimation of cloud-based streaming services

Quantitative quality estimation of cloud-based streaming services

Computer Communications

Research paper thumbnail of The use of big data analytics to predict the foreign exchange rate based on public media: A machine-learning experiment

The use of big data analytics to predict the foreign exchange rate based on public media: A machine-learning experiment

IT Professional

Research paper thumbnail of The Evolving Role of IT Departments in Digital Transformation

Sustainability

In the digital era, organizations are increasingly tasked with creating and utilizing new content... more In the digital era, organizations are increasingly tasked with creating and utilizing new content, applications, and/or services through the use of advanced information and communication technologies (ICT) to sustain a competitive advantage. Indeed, sustainability is now an embedded and overarching feature of organizations’ strategic planning. Research has shown that information technology (IT) departments are vital to organizations’ digital transformation. However, the role of IT departments in non-ICT-oriented organizations undergoing digital transformation has yet to be explored. Our study reveals that although the IT departments of non-ICT-oriented organizations play an important and proactive role in the early stages of organizational transformation and a dominant role in developing ICT capabilities, they will be unable to assume a leadership role within the organizations after transformation is complete.

Research paper thumbnail of Challenges in Creating Hybrid Professionalism Knowledge

Challenges in Creating Hybrid Professionalism Knowledge

IT Professional

Research paper thumbnail of Outlier detection in the concept drifting environment

Outlier detection in the concept drifting environment

2016 International Joint Conference on Neural Networks (IJCNN), 2016

Research paper thumbnail of The Business Process Investigation in the Perspective of Customer Value

The Business Process Investigation in the Perspective of Customer Value

International Conference on Electronic Business, 2004

Research paper thumbnail of The Layered Feed-Forward Neural Networks and Its Rule Extraction

The Layered Feed-Forward Neural Networks and Its Rule Extraction

Lecture Notes in Computer Science, 2004

Research paper thumbnail of The AI Tech-Stack Model

The AI Tech-Stack Model

Communications of the ACM

Management and technology challenges of AI-enabled application projects.

Research paper thumbnail of The Cramming, Softening and Integrating Learning Algorithm with Parametric ReLU Activation Function for Binary Input/Output Problems

The Cramming, Softening and Integrating Learning Algorithm with Parametric ReLU Activation Function for Binary Input/Output Problems

2019 International Joint Conference on Neural Networks (IJCNN), 2019

Rare Artificial Neural Networks studies address simultaneously the challenges of (1) systematical... more Rare Artificial Neural Networks studies address simultaneously the challenges of (1) systematically adjusting the amount of used hidden layer nodes within the learning process, (2) adopting Parametric ReLU activation function instead of tanh function for fast learning, and (3) guaranteeing learning all training data. This study will address these challenges through deriving the CSI (Cramming, Softening and Integrating) learning algorithm for the single-hidden layer feed-forward neural networks with the binary input/output and making the technical justification. To further verify the proposed learning algorithm, this study conducts an empirical experiment using SPECT heart diagnosis data set from UCI Machine Learning repository. The learning algorithm is implemented via the advanced TensorFlow and GPU.

Research paper thumbnail of The ICT predicament of new ICT-enabled service

ArXiv, 2015

The advancement of information and communication technologies (ICT) has triggered many ICT-enable... more The advancement of information and communication technologies (ICT) has triggered many ICT-enabled services. Regarding this service, the complementary ICT system involves with customers' devices, industry-wide ICT development and nation-wide ICT infrastructure, which are difficult for any individual organization to control. The ICT predicament is the phenomenon that the complementary ICT system is inferior in delivering the promised service quality of new ICT-enabled service. With the ICT predicament, companies face the decision-making dilemma in launching the new service or postponing the launch. This study proposes a process to resolve the decision-making dilemma regarding the ICT predicament.

Research paper thumbnail of Using Virtual Reality for Museum Exhibitions: The Effects of Attention and Engagement for National Palace Museum

Using Virtual Reality for Museum Exhibitions: The Effects of Attention and Engagement for National Palace Museum

Information technologies provide important opportunities for museums to create more engaging visi... more Information technologies provide important opportunities for museums to create more engaging visitor experience. This study collaborates with National Palace Museum and focuses on investigating whether virtual reality is an effective communication medium for museum exhibition. We compare three communication mediums-- video, website and virtual reality and investigate their impacts on user experience, in terms of engagement and attention. We are also interested in the role of users’ personality and their prior experience of communication media in moderating these effects. The results of findings show that virtual reality can catch more attention from user, and attention enforces their engagement, which leads to better user experience. Furthermore, personality and relative experience are both insignificant in mediating the relationship between engagement and visit intention, but we found that when using communication medium, extravert can have higher attention than the other users, which deserve more study in the future

Research paper thumbnail of How Design Features Lead to Visitors' Visit Intention through Virtual Reality Experience: The Case of National Palace Museum

How Design Features Lead to Visitors' Visit Intention through Virtual Reality Experience: The Case of National Palace Museum

The growing number of museums attempts to apply virtual reality technology to enhance the visitor... more The growing number of museums attempts to apply virtual reality technology to enhance the visitors’ experience, and significant research effort has been also made in the area of modeling the relics in VR. However, the studies on the real effects of the VR exhibition are less. In this paper, we draw upon the stimulus–organism–response (S- O-R) framework to theorize how system design features stimuli make an effect on the visitors’ experience of National Palace Museum’s VR exhibition, which in turn impacts the visit intention. Three types of design features are examined: interactivity, vividness, and realism. Additionally, we are interested in two types of VR experience: immersion and involvement. Furthermore, the study compares the VR effects across different kinds of relics (artifact, painting and calligraphy). The results of findings show that all of design features impact on immersion level for all relic types. Only for the calligraphy, the vividness impact on the involvement level, and only for the artifact, the interactivity impact on the involvement level

Research paper thumbnail of The Mathematical Programming and the Rule Ex-traction from Layered Feed-forward Neural Net-works

The mathematical programming analysis, instead of a data analysis, is proposed for identifying th... more The mathematical programming analysis, instead of a data analysis, is proposed for identifying the convex polyhedron associated with each rule. The area depicted in the rule premise covers a convex polyhedron in the input space, and the adopted approximation function for the output value is a multivariate polynomial function of x, the outside stimulus input. Moreover, the mathematical programming analysis is proposed for examining the extracted rules to explore features.

Research paper thumbnail of An improved back propagation neural network learning algorithm

An improved back propagation neural network learning algorithm

Artificial neural networks are dynamic computing systems which are made up of a number of simple,... more Artificial neural networks are dynamic computing systems which are made up of a number of simple, highly interconnected processing elements. They are designed to imitate some aspects of brain function such that they can self-organize, learn, and react. With these characteristics, artificial neural networks resemble those in the brain more closely than had previously been constructed. Yet the slowness of the convergence speed of learning (C.S.L.) undermines their applications to practical problems. Learning governed by an iterative optimization algorithm is a matter of searching for proper strengths of connections (weights) such that, at the end of learning, the network responds correctly to all training stimuli. For example, the Back Propagation learning algorithm (the BP algorithm) uses the gradient descent procedure to search for proper weights that minimize an objective function which is a measurement of the discrepancy between actual outputs and desired outputs with respect to all training stimuli. The research objective is to accelerate the C.S.L. We have studied two layer feed-forward artificial neural networks with one output node. We have achieved three things: extensive experimental observations of various 2-classes categorization problems with learning using the BP algorithm, explanations of these observations, and acceleration of the convergence speed of learning. Experimental observations show that there are some "attractors" in the landscape of the objective function that often interfere with the search process and cause the slowness of the C.S.L. These attractors are related to the existence of saddle stationary points. An improved version of the learning algorithm has been developed based on these observations. Experimental results obtained on a class of problems called parity problems have shown that this improved version is at least 15 times faster than using the conventional method which combined the momentum version of the gradient descent method and the adaptive stepsize technique.

Research paper thumbnail of The Research of Multi-Layer Topic Map Analysis using Co-word Analysis with Growing Hierarchical Self-organizing Map

International Journal of Digital Content Technology and its Applications, 2011

The purpose of this study was to propose a multi-layer topic map analysis using co-word analysis ... more The purpose of this study was to propose a multi-layer topic map analysis using co-word analysis of informetrics with Growing Hierarchical Self-Organizing Map (GHSOM). The topic map illustrated the delicate intertwining of subject areas and provided a more explicit illustration of the concepts within each subject area. We applied GHSOM, a text-mining Neural Networks tool, to obtain a hierarchical topic map. After taking up one example of altruism in evaluation, we suggest that topic map may disclose some important facts from a whole bunch of data.

Research paper thumbnail of A resistant learning procedure for coping with outliers

A resistant learning procedure for coping with outliers

Annals of Mathematics and Artificial Intelligence, 2009

In the context of resistant learning, outliers are the observations far away from the fitting fun... more In the context of resistant learning, outliers are the observations far away from the fitting function that is deduced from a subset of the given observations and whose form is adaptable during the process. This study presents a resistant learning procedure for coping with outliers via single-hidden layer feed-forward neural network (SLFN). The smallest trimmed sum of squared residuals principle

Research paper thumbnail of Artificial Intelligence in Smart Tourism: A Conceptual Framework

Smart tourism destination as: an innovative tourist destination, built on an infrastructure of st... more Smart tourism destination as: an innovative tourist destination, built on an infrastructure of state-of-the-art technology guaranteeing the sustainable development of tourist areas, accessible to everyone, which facilitates the visitor’s interaction with and integration into his or her surroundings, increases the quality of the experience at the destination, and improves residents’ quality of life. Lopez de Avila (2015). Smart tourism involves multiple components and layers of “smart” include (1) Smart Destinations which was special cases of smart cities integration of ICT’s into physical infrastructure, (2) Smart experience which specifically focus on technology-mediated tourism experience and their engagement through personalization, context-awareness and real-time monitoring, (3) Smart business refer to the complex business ecosystem that creates and supports the exchange of touristic resource and the co-creation of tourism experience. Gretzel et al, (2015). Smart tourism also cl...

Research paper thumbnail of The New Method of Feature Selection for Intradialytic Hypotension Prediction using Machine learning

The New Method of Feature Selection for Intradialytic Hypotension Prediction using Machine learning

2021 IEEE 3rd Eurasia Conference on Biomedical Engineering, Healthcare and Sustainability (ECBIOS)

Intradialytic hypotension (IDH) is a common complication and make patient get brain and heart dis... more Intradialytic hypotension (IDH) is a common complication and make patient get brain and heart disease. This study aimed to use the new method of feature selection to improve the performance of IDH prediction. We used feature importance by a different model to sort from high to low one by one, remove the lowest feature importance, and follow the rule again. Therefore, we obtain the combination of feature values that have the highest accuracy for predictive power. After applying REFCV with linear regression, the AUC of ROC increased to 0.966 (2-layers LSTM) and 0.965 (3-layers LSTM). The feature selection method we designed resulted in a better prediction ability in the model.

Research paper thumbnail of ANN Mechanism for Network Traffic Anomaly Detection in the Concept Drifting Environment

ANN Mechanism for Network Traffic Anomaly Detection in the Concept Drifting Environment

2018 IEEE Conference on Dependable and Secure Computing (DSC)

Research paper thumbnail of Managing Innovation and Cultural Management in the Digital Era

Managing Innovation and Cultural Management in the Digital Era

Research paper thumbnail of Quantitative quality estimation of cloud-based streaming services

Quantitative quality estimation of cloud-based streaming services

Computer Communications

Research paper thumbnail of The use of big data analytics to predict the foreign exchange rate based on public media: A machine-learning experiment

The use of big data analytics to predict the foreign exchange rate based on public media: A machine-learning experiment

IT Professional

Research paper thumbnail of The Evolving Role of IT Departments in Digital Transformation

Sustainability

In the digital era, organizations are increasingly tasked with creating and utilizing new content... more In the digital era, organizations are increasingly tasked with creating and utilizing new content, applications, and/or services through the use of advanced information and communication technologies (ICT) to sustain a competitive advantage. Indeed, sustainability is now an embedded and overarching feature of organizations’ strategic planning. Research has shown that information technology (IT) departments are vital to organizations’ digital transformation. However, the role of IT departments in non-ICT-oriented organizations undergoing digital transformation has yet to be explored. Our study reveals that although the IT departments of non-ICT-oriented organizations play an important and proactive role in the early stages of organizational transformation and a dominant role in developing ICT capabilities, they will be unable to assume a leadership role within the organizations after transformation is complete.

Research paper thumbnail of Challenges in Creating Hybrid Professionalism Knowledge

Challenges in Creating Hybrid Professionalism Knowledge

IT Professional

Research paper thumbnail of Outlier detection in the concept drifting environment

Outlier detection in the concept drifting environment

2016 International Joint Conference on Neural Networks (IJCNN), 2016

Research paper thumbnail of The Business Process Investigation in the Perspective of Customer Value

The Business Process Investigation in the Perspective of Customer Value

International Conference on Electronic Business, 2004

Research paper thumbnail of The Layered Feed-Forward Neural Networks and Its Rule Extraction

The Layered Feed-Forward Neural Networks and Its Rule Extraction

Lecture Notes in Computer Science, 2004