Search

Facial expression recognition (FER) is essential for discerning human emotions and is applied extensively in big data analytics, healthcare, security, and user experience enhancement. This paper presents an empirical study that evaluates four existing deep learning models—VGG16, DenseNet, ResNet50, and GoogLeNet—utilizing the Facial Expression Recognition 2013 (FER2013) dataset. The dataset contains seven distinct emotional expressions: angry, disgust, fear, happy, neutral, sad, and surprise. Each model underwent rigorous assessment based on metrics including test accuracy, training duration, and weight file size to test their effectiveness in FER tasks. ResNet50 emerged as the top performer with a test accuracy of 69.46%, leveraging its residual learning architecture to effectively address challenges inherent in training deep neural networks. Conversely, GoogLeNet exhibited the lowest test accuracy among the models, suggesting potential architectural constraints in FER applications. VGG16, while competitive in accuracy, demonstrated lengthier training times and a larger weight file size (512MB), highlighting the inherent balance between model complexity and computational efficiency.

<jats:p>Facial expression recognition (FER) is essential for discerning human emotions and is applied extensively in big data analytics, healthcare, security, and user experience enhancement. This study presents a comprehensive evaluation of ten state-of-the-art deep learning models—VGG16, VGG19, ResNet50, ResNet101, DenseNet, GoogLeNet V1, MobileNet V1, EfficientNet V2, ShuffleNet V2, and RepVGG—on the task of facial expression recognition using the FER2013 dataset. Key performance metrics, including test accuracy, training time, and weight file size, were analyzed to assess the learning efficiency, generalization capabilities, and architectural innovations of each model. EfficientNet V2 and ResNet50 emerged as top performers, achieving high accuracy and stable convergence using compound scaling and residual connections, enabling them to capture complex emotional features with minimal overfitting. DenseNet, GoogLeNet V1, and RepVGG also demonstrated strong performance, leveraging dense connectivity, inception modules, and re-parameterization techniques, though they exhibited slower initial convergence. In contrast, lightweight models such as MobileNet V1 and ShuffleNet V2, while excelling in computational efficiency, faced limitations in accuracy, particularly in challenging emotion categories like “fear” and “disgust”. The results highlight the critical trade-offs between computational efficiency and predictive accuracy, emphasizing the importance of selecting appropriate architecture based on application-specific requirements. This research contributes to ongoing advancements in deep learning, particularly in domains such as facial expression recognition, where capturing subtle and complex patterns is essential for high-performance outcomes.</jats:p>

<jats:p>Causal machine learning is an approach that combines causal inference and machine learning to understand and utilize causal relationships in data. In current research and applications, traditional machine learning and deep learning models always focus on prediction and pattern recognition. In contrast, causal machine learning goes a step further by revealing causal relationships between different variables. We explore a novel concept called Double Machine Learning that embraces causal machine learning in this research. The core goal is to select independent variables from a gesture identification problem that are causally related to final gesture results. This selection allows us to classify and analyze gestures more efficiently, thereby improving models’ performance and interpretability. Compared to commonly used feature selection methods such as Variance Threshold, Select From Model, Principal Component Analysis, Least Absolute Shrinkage and Selection Operator, Artificial Neural Network, and TabNet, Double Machine Learning methods focus more on causal relationships between variables rather than correlations. Our research shows that variables selected using the Double Machine Learning method perform well under different classification models, with final results significantly better than those of traditional methods. This novel Double Machine Learning-based approach offers researchers a valuable perspective for feature selection and model construction. It enhances the model’s ability to uncover causal relationships within complex data. Variables with causal significance can be more informative than those with only correlative significance, thus improving overall prediction performance and reliability.</jats:p>

It has been previously demonstrated that stochastic volatility emerges as the gauge field necessary to restore local symmetry under changes in stock prices in the Black–Scholes (BS) equation. When this occurs, a Merton–Garman-like equation emerges. From the perspective of manifolds, this means that the Black–Scholes and Merton–Garman (MG) equations can be considered locally equivalent. In this scenario, the MG Hamiltonian is a special case of a more general Hamiltonian, here referred to as the gauge Hamiltonian. We then show that the gauge character of volatility implies a specific functional relationship between stock prices and volatility. The connection between stock prices and volatility is a powerful tool for improving volatility estimations in the stock market, which is a key ingredient for investors to make good decisions. Finally, we define an extended version of the martingale condition, defined for the gauge Hamiltonian.

This paper examines the evolving trends in Chinese student mobility to Thailand, highlighting three distinct phases shaped by changes in the higher education: the dominance of Thai language programmes (1990–2010), the rise of business and international programmes (2010–2020), and the increasing preference for graduate studies (2020 onwards). By analysing the economic, cultural, and institutional factors facilitating these shifts, this paper positions Thailand as an emerging alternative study destination for Chinese students. It highlights the significance of this migration within the context of Thailand’s declining fertility rate and labour shortages, focusing on how Thai universities have adapted through active recruitment strategies targeting Chinese students. This paper also addresses the push and pull factors underpinning this migration and the pursuit of alternative educational pathways among Chinese youth. Additionally, it explores the strategic role of Sino-Thai collaborations under the BRI and their broader implications for educational mobility and economic ties.

The objective is to assess whether the extent to which employee resilience and organizational culture would be significantly related to and statistically predict the three facets of employee work engagement. Resilience was measured by four facets (Determination, Endurance, Adaptability, and Recuperability); and Organization Culture was measured for three types (Bureaucratic, Innovative, and Supportive). The dependent measures were the three facets of Work Engagement (Cognitive, Emotional, and Physical). This research by questionnaire was conducted in 2023. The questionnaires completed by 316 full-time workers revealed that all four facets of employee resilience had significant positive correlations with all three types of work engagement. Also, all three facets of work engagement were significantly higher in Innovative and Supportive cultures compared to Bureaucratic cultures. The regression analyses performed showed that the resilience factors of Determination and Adaptability were strong positive predictors of all three facets of work engagement. Furthermore, Innovative culture had additional positive effects on all three facets of work engagement; while Supportive culture had an additional positive effect on Emotional Work Engagement. The implications of the results for management are also discussed in this paper.

Macau Special Administration Region (Macau SAR) is in the process of revising legislation concerning special and inclusive education. While the institutional discourse revolves around establishing inclusive education, it is unclear as to how the proposed changes will enable or depress this from occurring. This research, therefore, examined teachers� attitudes towards inclusion as an indication of how well the new legislation may be received. Specifically, it investigated the interplay between 508 teachers working in private schools in Macau, that identified themselves as being inclusive schools, and teachers� sentiments and attitudes towards the acceptance of inclusion and the role that Confucian values might play in shaping these attitudes. Discussion focusses on four key outcomes that need to be addressed if a significant improvement in including all children in regular schools in Macau is to be achieved. These include the need (1) to clarify the concept of inclusion at government, school, and teacher levels as it currently has ambiguous meaning; (2) to provide teachers with more opportunities to have systematic contacts with students with SEN, as this is crucial to improving their sentiments and attitudes toward people with disability; (3) to provide professional learning about inclusive education with better partnerships between teacher education institutions and schools to bridge theory and practice; and (4) to review the hidden influence of the subtle levels of time-honoured Confucian beliefs in Macau, which are not manifest nor easily detected but possibly have a deep impact on day-to-day practices.

Examining why individuals engage in gambling is important in understanding the development of gambling behaviors. Numerous studies have investigated the underlying factor of gambling motivation. However, there is a dearth of evidence showing the latent dimensions of gambling motivation among individuals who are exposed to gambling in daily basis (i.e., casino employees). To address this gap, 817 casino employees were administered the Chinese version of the Gambling Motivation Scale (GMS) and other related measures. Results revealed that of the four models tested, a first-order model with seven factors achieved better fit in contrast to all other models. The seven factors include intrinsic motivation (IM) for knowledge, IM for accomplishment, IM for stimulation, extrinsic motivation (EM) due to identified regulation, EM due to introjected regulation, EM due to external regulation, and amotivation. However, the seven-factor model did not reach the conventional fit indices for good fit. After some post hoc modifications, the revised model achieved adequate fit. Moreover, the revised seven factors were related to frequency of gambling and amount spent for gambling. Generally, results showed that modified GMS with seven factors can be used with Chinese population, more specifically with Chinese casino employees.

PDF | Purpose Whilst the majority of academic studies have focused on the for-profit business-to-consumer type of sharing economy, the community-based... | Find, read and cite all the research you need on ResearchGate

Robotics are being used in the intervention with children with Autism Spectrum Disorder (ASD) in many places and already for many years. Many robots were developed and different studies are being made in order to evaluate its effectiveness. “Socially Assistive Robotics” is shown to be effective in different areas mainly in social and emotional development. Milo, a robot developed by a team led by Richard Margolin for the Robots4Autism program (RoboKind, 2020), is one of the robots whose use is reported to be successful. In Macao there is no report of studies or experiences on the use of robots in the intervention with children with ASD. In a collaboration between the Macao Science Centre, the Macao Autism Association (MAA) and the University of Saint Joseph, an exploratory study was developed to understand the applicability of Milo to the work with children with ASD in Macao. The study showed that the robot is able to facilitate social and emotional competences of children with ASD. However, several limitations including language, cultural differences, the inexperienced facilitators and the level of sessions are too simple for the participants to be aware of that may affect the effectiveness of the intervention. It is important to show that the adoption of Milo in Macao for intervening children with ASD can be further implemented, with better practical solutions.

This study examines how specific variables such as age, first language, nationality, school grade and socioeconomic status (SES) affect the comprehensibility of second language (L2) speech in 92 second/non-native language learners. Comprehensibility refers to the degree of speech understanding. Fluency, rhythm, grammatical features and word stressing are concurrent factors for the listening comprehension (and the listener comprehensibility) mainly in L2 context. Research evidence focused the quality and differences of speech samples produced by the L2 learners and the comprehensibility rated by native speakers. In reverse scenario there is less evidence on the judgment of L2 learners for speech samples produced by native speakers. In this study we analysed if the comprehensibility ability of 92 young Portuguese L2 learners differ in the following conditions: age, nationality, home language, school grade, proficiency and socioeconomic status. Speech (one text) was recorded by a native speaker and was judged by L2 speakers using 1-5 Likert scale for comprehension difficulty. Main results showed that neither age nor home language had influence for comprehensibility, but socioeconomic, nationality and grades accounted for statistical differences between the groups tested. Also, data suggested that phonetic features are more likely important for the beginner in second language learning compared to the semantic features of speech that heavily depend on vocabulary domain.

O estudo das estratégias que as crianças usam na leitura de palavras e a análise dos erros que tipicamente ocorrem ao longo do processo de aprendizagem numa dada ortografia são da maior importância para a compreensão do processo de aprendizagem da leitura. O objectivo deste trabalho foi perceber como se processa a aquisição da leitura no início do ensino fundamental no português europeu. Procurou-se saber se haveria diferenças na frequência de erros fonológicos e lexicais e no padrão de erros fonológicos entre os dois primeiros anos do ensino fundamental. Participaram 175 crianças do 1º ano e 137 do 2º ano de seis escolas. Foi aplicada uma prova de leitura oral de palavras. Os erros foram categorizados em fonológicos, com diversas subcategorias, e lexicais. Encontraram-se diferenças na frequência dos erros fonológicos e lexicais entre os dois anos, assim como nos subtipos de erros fonológicos que tipicamente ocorreram. Os erros de substituição foram os mais frequentes, tendo ocorrido mais nas consoantes e nos dígrafos. Seguiram-se os erros de adição e de supressão, que ocorreram sobretudo em sílabas complexas. Tais resultados, contribuindo para uma caracterização dos erros típicos nessa fase de aprendizagem, revelam-se um instrumento importante na detecção de dificuldades precoces na aprendizagem da leitura e na adequação de estratégias de ensino-aprendizagem. Este estudo permite uma melhor compreensão dos processos usados pelas crianças na resolução dos problemas que as características da língua portuguesa lhes coloca, assim como poderá permitir uma intervenção educativa que conduza a um maior sucesso na aprendizagem da leitura.Leitores principiantes; Aprendizagem; Leitura de palavras; Padrão de erros