34254 research outputs found
Sort by
Research on the Application and Performance Analysis of Heuristic Algorithms in Large Language Model Prompt Optimization
No full textIn recent years, the development of Large Language Models (LLMs) in AI has gained significant attention. These models excel in various Natural Language Processing tasks like text generation, translation, and sentiment analysis. Their optimization focuses mainly on two areas: Fine-tuning and Prompt Engineering. Fine-tuning improves model performance but requires extensive resources, while Prompt Engineering involves crafting input prompts for better efficiency and accuracy, demanding substantial expertise.
A new study(Guo et al., 2023) employed optimization algorithms to automate Prompt Engineering, aiming to reduce professional reliance, lower costs, and increase efficiency. This research used heuristic algorithms like Particle Swarm Optimization (PSO) and Simulated Annealing (SA) for LLM prompt optimization. The study compared these algorithms' effectiveness in Prompt Engineering, demonstrating PSO's strength in enhancing prompt diversity and SA's adaptability across various tasks.
We conducted a series of experiments using different datasets for various NLP tasks, such as Language Understanding, Text Summarization, and Text Simplification. These experiments aimed to verify the effectiveness of different optimization methods. The results showed that both PSO and SA have their advantages in enhancing prompt quality and overall LLM performance. In particular, PSO excels in enhancing the diversity and quality of prompts, while SA demonstrates its flexibility and adaptability across different tasks and contexts.
These research findings not only provide new methods and perspectives for the optimization of LLMs but also pave new paths for future research and applications in artificial intelligence, especially in the rapidly evolving field of natural language processing. Through these advanced optimization techniques, we can expect LLMs to play a more significant role in various practical applications in the future, thereby driving the development of the entire field of artificial intelligence
First-principles study on the topology of two-dimensional perovskite-type oxides
No full textTwo-dimensional (2D) non-van der Waals (vdW) ABO3 materials derived from the perovskite crystal structure have sparked significant attention recently due to their distinctive electronic and topological properties. However, there has been limited exploration of intriguing topological phases such as quantum spin Hall effects (QSH) and quantum-anomalous Hall effects (QAH) arising from the interplay between magnetism and spin-orbit coupling in these non-vdW 2D ABO3 materials. This dissertation delves into the novel nontrivial topological quantum phenomenon, QSH, and magnetic topological properties in 2D ABO3 materials predicted using first-principles calculations.
Firstly, we present the published work [Arokiasamy et al. CJP 86 (December 2023): 242\ue254] where we studied the quantum spin Hall and Rashba effects in 2D ilmenene oxides, ABO3 (A = Au, Ag, or Cu; and B = Bi, Sb, or As) pristine and their corresponding Janus structures. Interestingly, out of the 27 2D compounds examined in this study, two pristine and seven Janus compounds were predicted to have topological phases using the HSE06 hybrid functional since GGA underestimates electronic bandgaps. Furthermore, we chose AuBiO3 and Au0.5Ag0.5BiO3 to study the topology of a pristine material and its effect on doping. The non-trivial topology is confirmed by their Z2 invariance and conducting edge states under the hybrid functional approach. Moreover, we found van Hove singularities in Au0.5Ag0.5BiO3 and Cu0.5Ag0.5BiO3 near the Fermi level, suggesting the coexistence of superconductivity and nontrivial topology. The isotropic Rashba spin-splitting is studied in detail for Au0.5Ag0.5BiO3.
Subsequently, we studied hematene-type binary oxides and perovskites, which have recently gained significant research interest due to their potential for use in various devices. Herein, through first-principles calculations, we systematically examine the electronic, magnetic, and topological properties of structurally stable pristine 2D-hematene B2O3 (B = V, Nb, or Ta) and substitutionally doped 2D ABO3 (A=As, Sb, or Bi, and B = V, Nb, or Ta) perovskite structures. Interestingly, the pristine 2D-hematene Nb2O3 and Ta2O3 exhibit a topologically non-trivial phase that hosts the quantum spin Hall effect. Upon substituting one B atom with a pnictogen A atom, these compounds transform into an ABO3 perovskite-type oxide, exhibiting magnetic topological phases in its ferromagnetic ground state when spin-polarization and spin-orbit coupling are included. Without spin-orbit coupling, it demonstrates a Weyl ferromagnetic phase. The calculated Chern number (C) for the magnetic 2D ABO3 (A = As, Sb, or Bi, B = Nb or Ta) confirms their topologically non-trivial phases. The nontrivial Berry curvature and the potential topological edge states reveal the existence of magnetic topological properties in AsNbO3. This study identifies topological phases in 2D pristine B2O3 and substitutionally doped ABO3 materials that enable applications in spintronics, photonics, magnonics, memory, and nanoelectronics
Analysis of Mortgage Customers' Decision to Purchase Financial Products: A Case Study of E Bank
No full textIn recent years, the banking industry has faced changes in both the financial environment and regulatory requirements. Maximizing customer value and increasing income from wealth management services has become increasingly crucial in enhancing overall profitability of banks. This study analyzes and explores the factors that influence mortgage customers to make subsequent financial product purchases, and uses data from E Bank as the research sample. The research methodology includes a literature review, descriptive statistics, chi-square tests, Pearson correlation coefficients, and logistic regression analysis. Empirical findings revealed key elements regarding the correlation between mortgage cases and successful financial product transactions, thus allowing for a deeper understanding of mortgage customer behaviour. The results of this study will assist banks in formulating the development of their mortgage business by identifying which customer groups to focus more credit resources on, thereby creating higher commercial benefits
A Logic Implication Based On-Line Test and Protection Method for Approximate Logic Circuits
No full text\ue3\ue3Compared to traditional circuits, approximate circuits exhibit significant advantages in terms of performance, cost, and complexity. However, the discussion regarding their reliability is different from testing and fault tolerance from traditional circuits. Traditional testing and protection methods include Triple Modular Redundancy (TMR) and Built-In Self-Test (BIST). TMR is widely used in FPGA fault-tolerant design, testing, and protecting logic circuit outputs with threefold redundant areas. BIST includes both timing-based periodic testing and Scan-BIST types.
\ue3\ue3While these methods can be applied to testing and protection in traditional circuits, the consideration of acceptable and unacceptable faults in approximate circuits necessitates a new online testing method to accurately distinguish between these two types of faults.
\ue3\ue3This thesis proposes an on-line testing method utilizing implication, a method previously used for testing precise circuits, to test approximate logic circuits. By distinguishing between acceptable and unacceptable faults, it further filters and reduces the additional cost incurred due to implication. This approach aims to achieve cost savings while specifically detecting unacceptable faults.
\ue3\ue3Through the extension of the N-site hierarchy, a 100% detection rate for nonredundant faults(nR fault) is achieved while ignoring approximate redundant faults(axR fault). Additionally, this fault detection method is extended to design a fault-tolerant protection method for approximate logic circuits, exploring its cost-effectiveness further
An Analysis of Taiwan Investors' Behavior in the Context of High-Yield Financial Products: A Study of S Commercial Bank.
No full textThe primary objective of this research is to understand the investment behaviors of customers at a commercial bank, focusing on high-yield financial products. The study examines various client personality traits\ue2age, gender, occupation, educational level\ue2and financial product characteristics such as investment type, holding period, and performance. The goal is to profile customers of high-yield financial products and anticipate their future needs, guiding wealth management practices in banking.
The sample includes customers who invested in high-yield financial products, particularly mutual funds with an annual yield of 8% or higher, from January 1, 2015, to December 31, 2022. The data covers purchases, redemptions, and systematic investment plans, including various currencies and funds.
In-depth analysis reveals a correlation between customer characteristics and returns on high-yield mutual funds. It underscores that product choice and holding duration significantly impact returns. The study identifies key client segments, such as house managers and retirees, for high-yield financial products. Therefore, offering diverse products and integrated wealth management solutions is vital for market competitiveness and service optimization. This research aims to provide effective wealth management strategies for banks and clients, fostering a mutually beneficial scenario in wealth management
Effect of \uce\ub1-MSH in fibroblast cells of Thyroid Eye Disease
No full textThyroid eye disease (TED) is an autoimmune disease involving chronic orbital soft tissue inflammation and extraocular muscle enlargement. The tridecapeptide \uce\ub1-melanocyte stimulating hormone (\uce\ub1-MSH) is a well-studied treatment for multiple inflammatory disorders. This in vitro study investigated effects of \uce\ub1-MSH treatment on TED pathological processes. Orbital adipose specimens were obtained from chronic TED patients and surgical adipose speciments without dysthyroid history as control. MTT and lactate dehydrogenase release assays were performed to evaluate \uce\ub1-MSH effects on cell viability and apoptosis. From patient with TED and healthy controls, we developed orbital fibroblast primary cultures to examine the influence of \uce\ub1-MSH on pro-inflammatory cytokine production induced by interleukin (IL)-1\uce\ub2 by conducting real-time reverse transcription-polymerase chain reaction (PCR) analysis and western blotting. Immunofluorescence staining assay and quantitative PCR was performed to determine the expression level of proopiomelanocortin (POMC), the upstream neuropeptide of \uce\ub1-MSH in TED patients and non-TED controls. Treatment with non-cytotoxic concentrations of \uce\ub1-MSH resulted in the dose-dependent inhibition of inflammatory cytokines induced by IL-1\uce\ub2, including IL-6, IL-8, monocyte chemoattractant protein-1, intercellular adhesion molecule-1, and cyclooxygenase-2, at the levels of mRNA and protein expression (p < 0.05). Expression levels of POMC mRNA and protein were significantly higher in TED samples than non-TED controls (p < 0.05). Our findings reveal significant inhibition by \uce\ub1-MSH on inflammation, and POMC elevation was noted in chronic TED. To the best of our knowledge, this is the first in vitro study of the effect of \uce\ub1-MSH in TED. These findings indicate that POMC and \uce\ub1-MSH may have a role in immune regulation of TED and may be a potential therapeutic target
The Impact of ESG Ratings on Credit Risk Assessment of SMEs in Banks: Adjustment and Application of TESG Sustainable Development Indicators
No full textESG ratings are attracting increasing attention in the financial world. This study aims to explore the impact of ESG ratings on the credit risk of small and medium-sized enterprises, especially by adjusting the TESG sustainable development index to reflect the characteristics of small and medium-sized enterprises, and by calculating the correlation coefficient in the correlation analysis, this study attempts to determine whether there is a statistically significant association between ESG ratings and the credit risk levels of small and medium-sized enterprises. The research results highlight the importance of ESG ratings in credit risk assessment of SMEs and point out that specific ESG practices have potential value in improving corporate financial stability and bank credit evaluation. This finding provides a strong incentive for SMEs to improve their ESG performance, while also providing a new perspective for banks and financial institutions to integrate ESG factors into the credit risk assessment process
The exploration of determinants of university students' intention for internship
No full textThis qualitative research investigates the differences in student internship willingness across various academic disciplines in Taiwan using the Theory of Planned Behavior as the framework. Data were gathered from undergraduate students at National Sun Yat-sen University through online forums and focus groups. The findings revealed that internship willingness is influenced by both the students' academic disciplines and their perceived capabilities to engage in internships. Different career goals and employment direction clarity among students from various disciplines lead to diverse needs for career exploration, affecting their attitudes towards internship activities. Moreover, the perceived studying workload and availability of internship positions also play a role in shaping students' internship willingness. As a result, students from different academic backgrounds exhibit varying degrees of willingness to participate in internships
The Impact of Subsidies on Movie Theater Attendance: A Case Study of Taipei\ue2s Movie Theaters
No full textTo preserve Taiwan's national culture and protect it from the encroachment of dominant foreign cultures, the Taiwanese government has consistently implemented policies to support the domestic film industry, particularly through the protection and promotion of Taiwanese films. This support has taken the form of various policies, such as import quotas in the past and the current series of measures aimed at promoting and guiding the development of Taiwanese films. These policies have played a vital role in fostering the sustainable growth of the Taiwanese film industry. Among these measures, film subsidies have been a crucial source of funding for Taiwanese filmmakers. Therefore, this study focuses on the analysis of subsidies to investigate whether these supportive policies effectively achieve the objective of promoting Taiwanese films.
By examining data on the admissions of all Taiwanese films shown in Taipei cinemas between 2017 and 2022, along with information on subsidies over the years, this research employs multiple regression and the Entropy Balancing method in statistical matching to analyze the effectiveness of subsidy policies. Specifically, it explores the relationship between subsidies and the admissions for supported films.
The findings indicate that the distribution of subsidies indeed has a significant positive impact on the admissions of Taiwanese films, which differs from the results of previous subsidy-related studies
The Impact of the Housing Street Level Adjustment Rate on Housing Prices\uef\ubcTaking Kaohsiung City and Tainan City as an Example
No full textThe soaring housing prices in Taiwan have made housing justice a hot topic in recent years. In addition to the property hoarding tax, the housing tax system has long been criticized by many experts and scholars for the serious deviation of the current value of a house from the market price. The housing street level adjustment rate is an adjustment factor for the current value of a house, similar to the principles of land value taxation, and its necessity has been debated from various perspectives.
Therefore, this study aims to empirically analyze the causal relationship between the housing street level adjustment rate and housing prices in order to provide policy makers with relevant directions and improvement suggestions, and to achieve fairness in the domestic housing market and meet people's expectations for housing justice.
This study utilizes housing transaction data as the basis for housing characteristics variables in the 16 administrative districts of Kaohsiung City and Tainan City from 2016 to 2022. Additional variables such as location characteristics and spatial correlation are included. And a database of the housing street level adjustment rate is constructed using Geographic Information System (GIS). Through 3 econometric analysis methods, namely the hedonic price model, instrumental variable regression, and fixed effects model with similar individual centralization approach, the impact of the housing rate on housing prices is examined.
The results show that the the housing street level adjustment rate has a significant negative impact on housing prices in the models. Specifically, for every 10% increase in the housing street level adjustment rate, the average housing price decreases by 9.80%. This indicates a negative correlation between the housing street level adjustment rate and housing prices. Therefore, the housing street level adjustment rate, as an adjustment factor for property value, is necessary, and it is evident that the housing street level adjustment rate determined by the local government's Real Estate Assessment Committee has a significant influence on housing prices