34254 research outputs found
Sort by
Robust Asymmetric Multisupervised Discrete Hashing
No full textTypical hashing techniques primarily are aimed at training hash encodings with retrieval and classification capabilities. However, the optimization problem based on the label matrix of paired samples is a Quadratic Unconstrained Binary Optimization (QUBO) problem, which can be time-consuming or prone to substantial quantization errors when solved using current methods. Furthermore, during the data collection process, data corruption can occur, subsequently leading to model training biases. However, most hashing methods still used the L2 norms as the loss function, despite numerous studies confirming its sensitivity to noise. On the other hand, some hashing methods that used L2,p norms do not take into account the nonlinear structures of data and multisupervised information. Additionally, they involved one-step hashing, which might result in suboptimal performance.
This study proposes Robust Asymmetric Multisupervised Discrete Hashing (RAMDH). It first enhances the one-step learning of Discrete Hashing with Multiple Supervision using an asymmetric framework. The asymmetric framework enables closed-form solutions for each sub-optimization problem. Additionally, hash balance constraints, decorrelation constraints, and label regression are incorporated to enhance the quality of the hash codes. To address the issue of noise, this study employs a robust L2,p norms instead of L2 norms used in the two-step hash mapping matrix learning, subsequently turning the optimization problem into a nonconvex one. By rewriting the nonconvex expressions using a Reweighting (RW) optimization framework, the RW optimization framework can adapt itself to the current data. Ultimately, this study conducted experiments on four public datasets with three types of noise at five different noise intensities. The results demonstrated that the proposed method outperformed many existing supervised hashing methods when dealing with noisy data
Simulation of Three-Dimensional Photonic Liquid Crystals for Negative Refraction Analysis
No full textBlue phase liquid crystal is a three-dimensional periodic structure, recognized as a type of three-dimensional liquid photonic crystal. These liquid crystals are the phases that usually occur between the cholesteric and isotropic state. Blue phase liquid crystals can self-assemble to form three-dimensional crystals such as BPI with body-centered cubic structure and BPII with simple cubic structure, whose lattice constants are comparable to the wavelength of visible light. Due to their unique rapid electro-optic response and macroscopic optical anisotropy, blue phase liquid crystals have garnered significant academic interest for potential applications in optics. Furthermore, the three-dimensional periodic structure of blue phase liquid crystals bestows upon them a range of distinctive optical properties, making them suitable for optoelectronic applications. To comprehend the applicability of blue phase liquid crystals in optoelectronic components, an analysis of their photonic characteristics is important.
In this research, the finite-difference time-domain (FDTD) technique is utilized to systematically dissect the photonic band structure inherent in blue phase liquid crystals, employing a model predicated on a double twisted cylinder paradigm. Our simulations provide detailed insights into the equifrequency contour corresponding to the primary band of Blue Phase II. By invoking the conservation of momentum principle, it becomes feasible to forecast anomalous optical manifestations within blue phase liquid crystals. Intriguingly, within the equifrequency contour of Blue Phase II's primary band, we identify frequency lines with a distinct concave configuration in close proximity to the high-symmetry point M, signaling potential avenues for negative refraction phenomena.
In our simulation process, we methodically stacked Blue Phase II crystals to attain the requisite sample thickness and utilized a Gaussian Beam source for specific wavelength incidences. The emergence of negative refraction phenomena was discernibly captured through specialized receivers. Additionally, a point light source was harnessed to verify the converging characteristics of negative refraction. Concurrently, an in-depth analysis was undertaken to comprehend how liquid crystals with varied refractive indices demonstrate differing negative refraction traits. Upon nuanced elongation of Blue Phase II along the [101] crystallographic direction, we identified a deviation in negative refraction and its corresponding incident angles. These observations underscore our prowess in adjusting the angle of negative refraction via external fields, an accomplishment that remains elusive for traditional photonic crystals
Validation of the Calling Questionnaire
No full textThis study builds upon the findings of Chen, Ng, Chang, Chen, and Chen (2020) and collects a sample of 474 full-time employees. It confirms the good convergent validity of the sense of calling scale and demonstrates its significant incremental validity and explanatory power in relation to life meaning, job satisfaction, affective organizational commitment, turnover intention, life satisfaction, prosocial motivation and employee green behavior. Furthermore, it elucidates how calling influences individual behaviors or attitudes through different pathways.
Through hierarchical regression analysis and conditional indirect effect analysis, this research finds that life meaning mediates the relationship between individual calling and life satisfaction. Affective organizational commitment mediates the relationship between local calling and turnover intention. Prosocial motivation mediates the relationship between cosmopolitan calling and employee green behavior. Additionally, this study confirms that the conscientiousness personality trait moderates the relationship between cosmopolitan calling and employee green behavior through prosocial motivation. Finally, with regard to the research results, this study presents managerial implications
Single-Molecule Fluorescence Technique Identified Protein-DNA Interaction and Characterized the Biological Properties of Mitochondrial Helicase
No full textFluorescence is a sensitive and non-radioactive tool used in biophysics and biochemistry. There has been dramatic growth in the use of fluorescence for cellular and molecular imaging. Fluorescence detection can reveal the localization and interactions of intracellular molecules, sometimes at the level of single-molecule detection with good temporal and spatial resolution. In the first part of the study, the technique of protein-induced fluorescence enhancement or quenching (PIFE/PIFQ) was used to investigate interactions between protein and its DNA substrate labeled with Cy3 fluorophore. The dissociation equilibrium constant can be obtained by fitting bound fraction curve to binding model. RecA forms nucleoprotein filaments with single-stranded DNA during homologous recombination, which is an essential step for repairing DNA double-stranded breaks. The interactions between different RecA proteins with varying negative-charged residues in the C-terminal domain and single/double-stranded DNA was investigated here. We found that fewer negative-charged residues in the RecA C-terminal domain resulted in the formation of stable nucleoprotein filament. Hepatoma-derived growth factor (HDGF) has been reported that it can bind SMYD1 specifically. PIFQ was used to verify the sequence specificity of HDGF, PWWP and C140 sequentially, we found that C140 module can regulated the sequence-specific binding capability of HDGF on SMYD1.
In the second part, we used fluorescence resonance energy transfer (FRET) technique to investigated the DNA unwinding behaviors of mitochondrial DNA helicase Twinkle. A Cy3-Cy5 pair labeled substrate was used to probe distance change during Twinkle-mediated DNA unwinding process. The distance between the two fluorophores changed after helicase unwind dsDNA due to the change in persistent length of DNA substrate. The unwinding rate and behaviors of the helicase were analyzed. In addition, DNA substrate with consecutive mismatches were designed to examine whether the unwinding rate and behavior of Twinkle would be affected
The Impact of Movie Nostalgia on Brand Attachment and Brand Love\uef\ubcTake Top Gun Maverick as an Example
No full textThe modern entertainment industry plays a crucial role in addressing the lifestyle stress and mental health challenges faced by contemporary individuals. Top Gun
Maverick as a sequel to Top Gun, not only achieved significant success at the Taiwanese box office in 2022 but also emerged as a global box office champion. Grounded in attachment theory, this research focuses on the nostalgia, brand attachment, and brand love of moviegoers, particularly those in Taiwan who have watched Top Gun Maverick. The study meticulously examines the impact of nostalgia on brand attachment and brand love, incorporating self-congruity as a moderating variable. This approach aims to provide an in-depth understanding of the emotional connections\ue2 viewers form during the film, offering valuable practical and academic insights.
Conducted through quantitative research methods, the study utilized an online survey to collect data from audiences who had watched the film Top Gun Maverick. A total of 558 questionnaires were collected, with 533 deemed as valid samples, resulting in an impressive effective response rate of 95.52%. Statistical analysis employing the SPSS PROCESS macro model 7 was conducted to examine moderated mediation effects. The results revealed that nostalgia exert direct positive influences on both brand attachment and brand Love. Further investigations demonstrated that brand attachment plays a fully mediating role between nostalgia and brand Love. Additionally, selfcongruity was identified as a significant moderator in the relationship between nostalgia and brand attachment
Study on the continuing concatenation problem of motion mixture for skeleton model
No full textThis paper proposes a model composed of a neural network, and uses the State Variational Autoencoder as data compression and induction. The Encoder in SVAE compresses the current pose and the next pose into features Vector (Latent vector), the decoder generates the next pose according to the current pose and feature vector. The State Network in the decoder receives the output-feature vector of the encoder, and then combines it with the current pose to generate a state vector after softmax normalization. The decoder generates the next pose according to the state vector and the input current pose and feature vector; the conversion between the current pose and the next pose has been compressed to the feature vector through the SVAE network, so the agent only needs to observe according to the current pose and the task goal , the output vector cooperates with the current attitude control SVAE decoder to generate the next attitude. One of the objectives of the experimental task is to use a small amount of data to generate a series of gestures that are far away from the flag to reach the flag, design a synthetic control network, train two different decoders with SVAE, and use an agent to control two different The decoder outputs pose synthesis and generated poses, and additional step conditions are added so that the quality of motion will not drop too much when the two networks synchronously synthesize more poses; another task goal is to combine two unconnected motions from different sources Data, also use the synthetic control network to route the agent to find the closest posture by itself, so as to facilitate the blending and generation of postures that are connected together in the middle, and at the same time define the important joint positions to generate corresponding actions, so that no more data is needed Annotations on , enable the agent to control the decoder to make desired actions. From the experimental results, it can be observed that in the experiment of reaching the flag, the synthetic control network is used to make the output posture more diverse, so that it can reach the flag faster, and because the step condition does not reduce the movement value too much, the experiment of linking movements finds similar postures to carry out Synthesized, but the synthesized posture still has unreasonable results, and a better solution needs to be found; Finally, we can see that the designed control network has also successfully avoided the mentioned small amount of data problems, initial skeleton problems and acyclic type data problems
Workplace Telepressure and Counterproductive Work Behaviors: Exploring the Mediating and Moderating Mechanisms
No full textIn recent years, there has been many studies on workplace telepressure, and some pointed out that workplace telepressure has a negative impact on employees\ue2 physical and mental health, and reduces their resilience. However, few studies pointed out the impact of workplace telepressure on employees\ue2 psychological state could pose threat to the organization. Three purposes are listed below. First, explore the relationship between employees' workplace telepressure and counterproductive work behaviors. Second, test the mediating effect of work anxiety between workplace telepressure and counterproductive work behaviors. Three, explore whether role breadth self-efficacy and emotion regulation will strengthen or weaken the impacts of workplace telepressure on job anxiety and counterproductive work behaviors. The study adopts a one-stage questionnaire survey. A survey is conducted among a sample of 220 employees in Taiwan. The results show that\uef\ubc
1. Interpersonally directed counterproductive work behaviors are positively related to workplace telepressure.
2. Workplace telepressure is positively related to job anxiety.
3. The relationship between workplace telepressure and interpersonally directed counterproductive work behaviors is mediated by job anxiety.
4. The relationship between workplace telepressure and organizationally directed counterproductive work behaviors is mediated by job anxiety
A Hierarchical Architecture for Multi-Agent Cooperative Systems
No full textIn this paper, we propose a method to apply hierarchical reinforcement learning to multi-agent cooperation. In A Hierarchical Architecture for Multi-Agent Cooperative Systems, only local environmental information is obtained, and the task is divided into two stages using hierarchical reinforcement learning, with the manager at the top deciding the main direction and splitting the main direction into smaller tasks for the executors at the bottom. The intrinsic reward mechanism of hierarchical reinforcement learning is divided into two types: one is based on the entropy of attention weights to calculate intrinsic rewards, and the other is based on the goal function and gradient direction to calculate intrinsic rewards.
In the paper, the effectiveness of the termination network is tested by comparing the entropy of attention weights as intrinsic reward, the target function and gradient direction as intrinsic reward, and only the final task completion as reward
Participatory modeling for adaptation to multiple interacting stressors: an exploration in Cigu coastal area
No full textWhile climate change adaptation captures the most attention, a place-centered adaptation planning requires consideration of multiple interacting stressors that local communities concern. This is the case for Cigu, a coastal area in Taiwan that contains rich and diverse ecosystems ranging from mangroves to salt fields, and from fish to birds (e.g., endangered black-faced spoonbill). It is also economically noteworthy for fish-pond aquaculture, oyster farming and ecotourism. However, decrease in sediment transport causes geomorphological change. Sand barrier islands of Cigu Lagoon have eroded, narrowed, lowered and moved landwards. Imported aquaculture products and increasing aquaculture costs threaten local livelihood. Large-scale photovoltaics are expected to bring major impacts to Cigu. And Cigu is a rural area with decreasing, aging population, becoming a \ue2super-aged society\ue2. These multiple stressors prompted a participatory modeling process designed, conducted and evaluated to engage diverse stakeholders for social learning. In this process, rich context was learned from the field. Stakeholders were identified and mapped using a stakeholder rainbow diagram and a power\ue2interest grid. Group Model Building scripts were used to organize meetings, facilitating participants to share their hopes and fears, to draw graphs over time of the variables they concerned, to draw causal loop diagram of the problem together, to quantify System Dynamic model, and to discuss scenarios and policy simulations. The results show that participatory modeling can support social learning for long-term adaptation. Participants have systems thinking perspectives, better understanding of the dynamics and closer relationships. This initial exploration may illuminate application of participatory modeling in adaptation planning and suggest future improvements
A Study on the Relationship among Consumer Value, Psychological Ownership, and Purchase Intention When Buying Experiential Products Using Different Payment Method
No full textBNPL (Buy Now Pay Later) payment service provided by the companies like Affirm or Afterpay, is become increasingly popular as an electronic payment in foreign countries. This thesis considers and compares BNPL and one-off full payment, and investigates whether consumers\ue2 use of the above two different payment methods for purchasing experience products will result in different consumer values, and how these consumer values affect consumers\ue2 perceived psychological ownership of the products. Additionally, this study investigates mental accounting effect that makes it easier for consumers to track their spending, affects consumers\ue2 perceived psychological ownership and purchase intention towards experience products. Relevant data was obtained through two sets of questionnaires with scenarios of using BNPL or one-off full payment to book a trip to Japan (the experience product). Partial least square method analysis on 134 and 113 survey returns reveals that consumer values differ according to the payment methods, and subsequently, these values impact consumers' perceived psychological ownership of the products. Moreover, in contrast to one-off full payment, mental accounting effect of BNPL significantly influences consumers' perceived psychological ownership of the products