As society advances, an increasing number of individuals spend significant time interacting with computers daily. To enhance the human-computer interaction experience, it has become crucial to augment the computer’s ability for emotion recognition. This capability holds excellent importance as machines become capable of responding to us in a more natural and contextually relevant manner, aligned with our current emotional states. Examples of such applications include caregiving and social robots. Accurate recognition of human emotions, followed by the ability to determine the most appropriate responses, can significantly enhance user experiences. The most commonly employed methods in emotion recognition include observing facial expressions, audio, and conversational content. The multi-modal emotion recognition lacks the explicit mapping relation between emotion state and audio and image features. This study proposes a fusion method for audio-visual emotion recognition. The audio and video data are preprocessed separately. The audio emotion features and visual expression features were then extracted using two distinct feature extractors. The audio emotion feature extractor, denoted as audio-net, employs a 2D CNN architecture capable of processing image-based Mel-spectrograms as input data. The facial expression feature extractor, visual-net, uses a 3D CNN architecture to process sequences of facial expression images. Fusing the visual and auditory features and enhancing feature correlation using the deep canonical correlation analysis (DCCA) method. This research uses the eNTERFACE05 dataset and reaches 89.13% accuracy in classifying emotions. The result shows that considering audio and facial features at the same time can the model better recognize the emotion people are having.

In the “Computer Programming (II)” course at National Taipei University of Technology, students submit their code for assignments, which undergo evaluation based on multiple test cases. To deter direct access to answers via test case results, the Assignment Submission System only reveals the execution outcome of the first test case and compile errors. It shows the serial numbers of passed or failed test cases without disclosing their details. However, this limited feedback poses challenges for effective debugging for the students. This study proposes a Student Program Analysis and Feedback System employing techniques like Spectrum-Based Fault Localization, Dynamic Slicing, GDB, and Valgrind. The system aims to provide constructive error feedback to aid students in rectifying their assignments efficiently, thereby reducing debugging time and effort.

Most of today's programming courses use online judge systems as course materials. With the increase of courses and people in the field of computer science and information engineering, the use of online judge systems is becoming more and more widespread, but simultaneously, there are more and more attacks on online judge systems. So how avoiding these attacks on online judge systems is becoming more and more important. This research studies and organizes these attack methods, and creates a threat model for the online judge system, to design code analysis rules and implement a code analysis tool. This tool can help developers analyze the existing online judge system to check whether the judicial system is at risk of being attacked and to deal with it as soon as possible to enhance the security of the judge system.

This paper extended Stacked Denoising Autoencoder to build a deep neural network which initialized the weight of neural network through the encoder's weight and used Dropout to reduce the error rate in fine-tuning stage. The neural network used the information of students in recent years as input data to train neural network, and predicted the possibility of dropout on the students during the semester. The prediction result can be used to counseling and warning students which be dropout likely and then reduced the unnecessary resource of school.

Accurate Human Epithelial-2 (HEp-2) cell image classification plays an important role in the diagnosis of many autoimmune diseases. However, the traditional approach requires experienced experts to artificially identify cell patterns, which extremely increases the workload and suffer from the subjective opinion of physician. To address it, we propose a very deep residual network (ResNet) based framework to automatically recognize HEp-2 cell via cross-modal transfer learning strategy. We adopt a residual network of 50 layers (ResNet-50) that are substantially deep to acquire rich and discriminative feature. Compared with typical convolutional network, the main characteristic of residual network lie in the introduction of residual connection, which can solve the degradation problem effectively. Also, we use a cross-modal transfer learning strategy by pre-training the model from a very similar dataset (from ICPR2012 to ICPR2016-Task1). Our proposed framework achieves an average class accuracy of 95.63% on ICPR2012 HEp-2 dataset and a mean class accuracy of 96.87% on ICPR2016-Task1 HEp-2 dataset, which outperforms the traditional methods.

In recent years, automated service composition has been a fervid research area in service computing. Within this area, service description plays a crucial role in terms of the development of a diverse number of automation schemes. In this paper, we provide an investigation and classification of the service description approaches that have been used in a diverse collection of automated service composition studies. To position the service description approaches used in automated service composition throughout the service description world and to clearly classify them, we propose five two-value dimensions. Using the proposed dimensions, we first perform a categorization and provide a simple introduction to current representative industrial service description standards. Subsequently, because we discovered that most of the studied automated composition approaches follow a tuple-based service description paradigm, an exhaustive classification and discussion of this paradigm is made, with the automated composition approaches adopting this paradigm as an example. Finally, we discuss issues that are currently relevant to the service description field and possible solutions. With this study, the reader can obtain a complete understanding of service description approaches used in automated service composition research, including their common formulation and assumptions.

This paper proposes an argumentation and negotiation mechanism for multi-agent systems. Through argumentations and negotiations, agents obtain more information on the topics of common interests or on those they have odds with. At the inception of the negotiation, agents can hardly understand completely the goals and beliefs other agents have toward related issues. Through argumentations and negotiations, the beliefs evolve, and agents will have better understanding about each other's target needs and preferences. During negotiations, agents can select the proposal that better suits other agents, further improving the chances for the agents to reach a consensus. Lastly, this paper illustrates our proposed methods through a simple course-scheduling negotiating system.

Project estimation based on function point or use case point (UCP) methods provide only fixed complexity grades which can not deal with the uncertain and imprecise conditions. This study, therefore, provides a fuzzy size estimation procedure for goal-driven use case model based on UCP using fuzzy theory. We propose a metric to calculate the unadjusted use case points of goal-driven use cases based on the relations between each use cases and goals with the fuzzy membership functions and the fuzzy rules. Furthermore, the technical and environmental factors are considered to calculate the use case points which can be used to estimate the implementation time and effort of the system under development. The proposed approach is illustrated by a benchmark problem domain of a meeting scheduler systems.

Service-oriented architecture implemented by Web Services is one of the most popular and promising software development paradigm that has brought some challenging research issues today. One of the most important issues is how to automate web service composition at design phase. Currently, there are many researchers concentrating on service composition problem that can be partitioned into three parts, dynamic workflow composition, QoS-aware, and transaction-aware service selection. This paper addresses the issue of automatic composing Web Services into an executable workflow not only according to user's functional requirements but also to their transactional properties and QoS characteristics. We propose an automatic composition approach through genetic algorithm to satisfy user's functional requirements, QoS criteria, and transactional requirements automatically at the same time. Experimental results are presented.

For the video surveillance system nowadays, identifying the color of certain footage is paramount. Every time when it comes to a crime scene, the police will be able to extract useful information from the surveillance cameras on the scene. Among that important information, “color” plays a major role and is not affected by the size, location, time or changes in shape of an object. Facts affecting the accuracy and efficiency of realtime color identification include the material and the parameter of camera lens, the parameters of infrared for the hardware part, and efficiency of software algorithms, accuracy and practical degree for the software part, and so forth. Although many methods using static analysis of color have been proposed, they cannot effectively solve the color recognition in the real-time video system. In this paper, we provide an approach using dynamic algorithm for the real-time video system. The methodologies include the reduction of color dimension, color transformation, color classification and real-time color recognition. Lastly, we provide the methods for effectively improving both the efficiency and accuracy of a real-time surveillance system.

RoboCup soccer game is a competition game. Robots need good strategy and planning ability on different conditions to win the game. If robots act by prior rules, they cannot gain ground. In this paper, we proposed a hybrid approach to implement our RoboCup agent. It can provide good strategy for robots planning base on all kinds of conditions and saving experience for reusing. Robots will grow up by our hybrid approach without prior defining knowledge and complex math basis. They just learn by saving experience. The robots don't only grow up but also avoid to making the same mistakes. And we show the effectiveness of the proposed method through implement and comparing with another learning approach

This paper presents an adaptive approach to address the goal evolution of the intelligent agent. When agents are initially created, they have some goals and few capabilities. These capabilities can perform some actions to satisfy their goals. They strive to adapt themselves to the low capabilities. Reinforcement learning method is used to the evolution of agent goal. An abstract agent programming language (3APL) is introduced to build the agent mental states. We propose reinforcement learning to refine the top-level goals. A robot soccer game is used to explain our approach. Moreover, we show how a refinement of the soccer player's mental state is derived from the evolving goals by reinforcement learning.

Online course management system (OCMS) mainly aids various events in online instructing, including testing, course discussion, assignment submission, and assignment grading. This paper is mainly designed basing on the study of completed OCMS of the past. Online assignment submission is prone to easy plagiarism, infecting the learning process of the students and interfering with their studies. In the past, using human power to inspect for plagiarism is very time-consuming. This research then is focused on allowing programming courses to employ procedures such as code standardization, textual analysis, structural analysis, and variable analysis, to evaluate and compare programming codes. We provide an intelligent agent as a daemon to analyze the program code for OCMS. Textually, we use document fingerprinting algorithm as a basis for text comparison; structurally, we utilize formal algebraic expression and dynamic control structure tree (DCS tree) to rebuild and evaluate the program structure; variable-wise, we not only record relevant information for each variable, but also analyze the programming structure where the variables are positioned. By applying a similarity measuring method, we output a similarity value for each program in the three aspects mentioned above. This research implements a convenient user interface that can be applied independently for assignment analyzation. Moreover, we have designed a set of application programming interface (API) that could be embedded into online course management systems.

This work presents a fuzzy mental model to address evolution of intelligent agents in virtual market. Agent fitness and fuzzy multi-criteria decision-making are proposed as evolution mechanisms, and fuzzy soft goals are introduced to facilitate the evolution process. Genetic programming operators are employed to restructure agents in the proposed multi-agent evolution cycle. We conduct a series of experiments to determine the most successful strategies and to see how and when these strategies evolve based on the context and the bidding stance of the agent's opponent.

Fuzzy theory is suitable to capture and analyze the informal requirements that are imprecise in nature, meanwhile, XML is emerging as one of the dominant data formats for data processing on the Internet. In this paper, we attempt to markup the fuzzy objects with XML, and provide conversion rules to define the mapping from fuzzy objects model and fuzzy objects specification into XML schema and XML documents, respectively.

In this paper, we have attempted a study of current approaches carried out in the confluence of the two technologies: fuzzy set theory and software engineering, that could provide a powerful tool for requirements engineering, formal specifications, software quality prediction, object-oriented modeling, and etc. Various requirements analysis and specifications modeling technologies that utilize fuzzy theory are identified, and works related to the use of fuzzy logic for predicting software quality are also outlined.

In this study, we have attempted a survey of current approaches carried out in the confluence of the two technologies, fuzzy set theory and object-oriented technology, that could provide a powerful tool for enhancing database management systems, software modeling, and knowledge representation in artificial intelligence (AI) systems. Possible types of fuzziness are discussed and key features related to different kinds of fuzzy software systems are also pinpointed. In a nutshell, fuzzy theory, as a modeling mechanism, is especially useful in tackling real world applications whose complexity demands are growing intensively. © 2001 John Wiley & Sons, Inc.

The focus of the paper is on the use of time Petri nets to serve as the verification mechanism for acquired scenarios. Use cases are used to elicit the user needs and to derive the scenarios. After specifying all possible scenarios, each of them can be transformed into its corresponding time Petri nets model (TPN). Through the analysis of these TPN models, wrong information and missing information in scenarios can be detected. The proposed approach is illustrated by a course registration problem domain.