5916 research outputs found
Sort by
Explanation-based negotiation protocol for nutrition virtual coaching
No full textPeople’s awareness about the importance of healthy lifestyles is rising. This opens new possibilities for personalized intelligent health and coaching applications. In particular, there is a need for more than simple recommendations and mechanistic interactions. Recent studies have identified nutrition virtual coaching systems (NVC) as a technological solution, possibly bridging technologies such as recommender, informative, persuasive, and argumentation systems. Enabling NVC to explain recommendations and discuss (argument) dietary solutions and alternative items or behaviors is crucial to improve the transparency of these applications and enhance user acceptability and retain their engagement. This study primarily focuses on virtual agents personalizing the generation of food recipes recommendation according to users’ allergies, eating habits, lifestyles, nutritional values, etc. Although the agent would nudge the user to consume healthier food, users may tend to object in favor of tastier food. To resolve this divergence, we propose a user-agent negotiation interacting over the revision of the recommendation (via feedback and explanations) or convincing (via explainable arguments) the user of its benefits and importance. Finally, the paper presents our initial findings on the acceptability and usability of such a system obtained via tests with real users. Our preliminary experimental results show that the majority of the participants appreciate the ability to express their feedback as well as receive explanations of the recommendations, while there is still room for improvement in the persuasiveness of the explanations.Italian Ministry for Universities and Research ; Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung ; Fonds National de la Recherche Luxembourg ; CHIST-ERA ; TÜBİTA
Assembly-based flood repair cost and time models for industrial buildings in Turkey
No full textFlooding is one of the most frequent hazard events significantly affecting the industry in Turkey, leading to severe economic losses and business disruptions. To predict the direct flood losses and business downtimes on the industry, this study proposes probabilistic cost and time models of repairing direct physical flood damage to industrial buildings in Turkey. Using field notes and literature reviews, a typical industrial building was first disassembled into a list of building components vulnerable to flooding. With a focus on façade and interior building components, the flood damage potential of each component was assessed at different flood depths and velocities. Damage state relationships were established for the building components to connect their damage levels under flood actions to their individual repair work. Using the assembly-based approach, costs and times of repairing each flood-damaged component and their associated variabilities were assembled in a probabilistic approach to develop total flood repair cost and time models. The results show that typical repair costs and times for facilities subjected to significant flood depths and velocities might reach substantial levels, up to 28% of building replacement costs and 165 working days on average. The proposed models systematically incorporate existing important uncertainties, therefore, provide reliable estimates.TÜBİTA
Reducing shape errors in the discrete dipole approximation using effective media
No full textThe discrete dipole approximation (DDA) simulates optical properties of particles with any given shape based on the volume discretization. These calculations cost a large amount of time and memory to achieve high accuracy, especially for particles with large sizes and complex geometric structures, such as mixed black-carbon aerosol particles. We systematically study the smoothing of the DDA discretization using the effective medium approximation (EMA) for boundary dipoles. This approach is tested for optical simulations of spheres and coated black-carbon (BC) aggregates, using the Lorenz-Mie and multiple-sphere T-Matrix as references. For spheres, EMA significantly improves the DDA accuracy of integral scattering quantities (up to 60 times), when the dipole size is only several times smaller than the sphere diameter. In these cases, the application of the EMA is often comparable to halving the dipole size in the original DDA, thus reducing the simulation time by about an order of magnitude for the same accuracy. For a coated BC model based on transmission electron microscope observations, the EMA (specifically, the Maxwell Garnett variant) significantly improves the accuracy when the dipole size is larger than ¼ of the monomer diameter. For instance, the relative error of extinction efficiency is reduced from 4.7% to 0.3% when the dipole size equals that of the spherical monomer. Moreover, the EMA-DDA achieves the accuracy of 1% for extinction, absorption, and scattering efficiencies using three times larger dipoles than that with the original DDA, corresponding to about 30 times faster simulations.National Natural Science Foundation of China ; Nanjing University of Information Science and Technology ; Graduate Research and Innovation Projects of Jiangsu ProvincePublisher versio
SGD with partial hessian for deep recommendation system optimization
No full textDue to the effectiveness of second-order algorithms in solving classical optimization problems, designing secondorder optimizers to train deep neural networks (DNNs) has attracted much research interest in recent years. However, because of the very high dimension of intermediate features in DNNs, it is difficult to directly compute and store the Hessian matrix for network optimization. Most of the previous secondorder methods approximate the Hessian information imprecisely, resulting in unstable performance. In this work, we propose a compound optimizer, which is a combination of a second-order optimizer with a precise partial Hessian matrix for updating channel-wise parameters and the first-order stochastic gradient descent (SGD) optimizer for updating the other parameters. We show that the associated Hessian matrices of channel-wise parameters are diagonal and can be extracted directly and precisely from Hessian-free methods. Our proposed method, namely SGD with Partial Hessian for Deep Recommendation System (SPHDRS), inherits the advantages of both first-order and secondorder optimizers. Compared with first-order optimizers, it adopts a certain amount of information from the Hessian matrix to assist optimization, while compared with the existing secondorder optimizers, it keeps the good generalization performance of first-order optimizers. Experiments on synthetic and real recommendation system tasks demonstrate the effectiveness of our proposed optimizer SPH-DRS
Exploring well-being dynamics across career stages: Challenges and opportunities for EFL teachers
No full textThis session delves into the dynamic nature of well-being among English as a Foreign Language (EFL) teachers, examining how it changes across various career stages. In the context of an ever-evolving educational landscape with fast-technology improvements and pandemics, understanding the well-being challenges and opportunities faced by EFL teachers is crucial for promoting their professional development, personal satisfaction, workload, student dynamics, institutional support, and personal growth. Combining extensive research and real-life experiences, this study explores the features of well-being transitions from novice to experienced EFL colleagues. The paper will shed light on the opportunities within these challenges. It will emphasize the importance of mentorship, professional development, and self-care practices at different career stages to enhance well-being. The role of educational institutions and policymakers in fostering a supportive environment for EFL teachers will also be
explored, ultimately benefiting both teachers and their students
A 3-DoF robotic platform for the rehabilitation and assessment of reaction time and balance skills of MS patients
No full textThe central nervous system (CNS) exploits anticipatory (APAs) and compensatory (CPAs) postural adjustments to maintain the balance. The postural adjustments comprising stability of the center of mass (CoM) and the pressure distribution of the body influence each other if there is a lack of performance in either of them. Any predictable or sudden perturbation may pave the way for the divergence of CoM from equilibrium and inhomogeneous pressure distribution of the body. Such a situation is often observed in the daily lives of Multiple Sclerosis (MS) patients due to their poor APAs and CPAs and induces their falls. The way of minimizing the risk of falls in neurological patients is by utilizing perturbation-based rehabilitation, as it is efficient in the recovery of the balance disorder. In light of the findings, we present the design, implementation, and experimental evaluation of a novel 3 DoF parallel manipulator to treat the balance disorder of MS. The robotic platform allows angular motion of the ankle based on its anthropomorphic freedom. Moreover, the end-effector endowed with upper and lower platforms is designed to evaluate both the pressure distribution of each foot and the CoM of the body, respectively. Data gathered from the platforms are utilized to both evaluate the performance of the patients and used in high-level control of the robotic platform to regulate the difficulty level of tasks. In this study, kinematic and dynamic analyses of the robot are derived and validated in the simulation environment. Low-level control of the first prototype is also successfully implemented through the PID controller. The capacity of each platform is evaluated with a set of experiments considering the assessment of pressure distribution and CoM of the foot-like objects on the end-effector. The experimental results indicate that such a system well-address the need for balance skill training and assessment through the APAs and CPAs.Publisher versio
On the density of branching Brownian motion
No full textWe consider a d-dimensional dyadic branching Brownian motion, and study the density of its support in the region where there is typically exponential growth of particles. Using geometric arguments and an extension of a previous result on the probability of absence of branching Brownian motion in linearly moving balls of fixed size, we obtain sharp asymptotic results on the covering radius of the support of branching Brownian motion, which is a measure of its density. As a corollary, we obtain large deviation estimates on the volume of the r(t)-enlargement of the support of branching Brownian motion when r(t) decays exponentially in time t. As a by-product, we obtain the lower tail asymptotics for the mass of branching Brownian motion falling in linearly moving balls of exponentially shrinking radius, which is of independent interest.Publisher versio
Ethnicity-related partner selection experiences among young adults from christian armenian families: A qualitative study in turkey
No full textIn this qualitative study partner selection experiences of young Armenian adults were examined using Ecological Systems Theory and Filter Theory as a theoretical basis. The literature review of partner selection was reviewed in three subtitles as theoretical foundations of partner selection, partner selection studies in Turkey, and intermarriage and intramarriage in Armenians. The study was conducted with 10 participants and data analysis was undertaken through thematic analysis using MAXQDA. This approach emerged nine themes and eight subthemes across four levels. Individuals’ Filters and Preferences about Partner Selection Level themes were as follows: 1. Choosing an Armenian partner is a priority, 2. Importance of the Partner’s Ethnic Identity is Highlighted for Long-Term Relationships, 3. Implementing other selection filters than ethnic identity, 4. Advising Armenian Partner to Children without Rigid Restrictions. Impact of Immediate Social Environment level emerged as Theme 5: Community disapproves and judges out-group marriages, 6. Friends tend to select Armenian partners w/out isolating others, and 7. Families implement and contain partner restriction, Social Connection Level theme revealed: 8. Strength of ties to Armenian Culture/Community promote deeper interiorizing of ingroup partner selection, Attitudes and Ideologies of the Culture Level emerged as 9. Concerns about sustaining culture/population promote in group partner selection. Themes and Subthemes were elaborated upon and discussed in the results and discussion sections providing comprehensive insight into the partner selection experiences of young Armenian Adults. Thematic analysis of the interviews exhibited the interaction of the larger social systems during the partner selection process of Armenian young adults. Thus, this study introduced an extended framework illuminating partner selection experiences among ethnic minorities in the Turkish context
Exploration with intrinsic motivation using object–action–outcome latent space
No full textOne effective approach for equipping artificial agents with sensorimotor skills is to use self-exploration. To do this efficiently is critical, as time and data collection are costly. In this study, we propose an exploration mechanism that blends action, object, and action outcome representations into a latent space, where local regions are formed to host forward model learning. The agent uses intrinsic motivation to select the forward model with the highest learning progress to adopt at a given exploration step. This parallels how infants learn, as high learning progress indicates that the learning problem is neither too easy nor too difficult in the selected region. The proposed approach is validated with a simulated robot in a table-top environment. The simulation scene comprises a robot and various objects, where the robot interacts with one of them each time using a set of parameterized actions and learns the outcomes of these interactions. With the proposed approach, the robot organizes its curriculum of learning as in existing intrinsic motivation approaches and outperforms them in learning speed. Moreover, the learning regime demonstrates features that partially match infant development; in particular, the proposed system learns to predict the outcomes of different skills in a staged manner