64 research outputs found
Batch Bayesian Learning of Large-Scale LS-SVMs Based on Low-rank Tensor Networks
No full textLeast Squares Support Vector Machines (LS-SVMs) are state-of-the-art learning algorithms that have been widely used for pattern recognition. The solution for an LS-SVM is found by solving a system of linear equations, which involves the computational complexity of O(N^3). When datasets get larger, solving LS-SVM problems with standard methods becomes burdensome or even unfeasible. The Tensor Train (TT) decomposition provides an approach to representing data in highly compressed formats without loss of accuracy. By converting vectors and matrices in the TT format, the storage and computational requirements can be greatly reduced. In this thesis, we develop a Bayesian learning method in the TT format to solve large-scale LS-SVM problems, which involves the computation of a matrix inverse. This method allows us to include the information we know about the model parameters in the prior distribution. As a result, we are able to obtain a probability distribution of the parameters, which enables us to construct confidence levels of the predictions. In the numerical experiment, we show that the developed method performs competitively with the current methods.Mechanical Engineering | Systems and Contro
Formation of nano-twinned structure in Ti3AlC2 induced by ion-irradiation
No full textIn the present study, a nano-twinned microstructure is found in ion irradiated-Ti3AlC2 by the characterizations of high-resolution transmission electron microscopy and electron diffraction, combined with grazing incidence X-ray diffraction. No obvious difference in chemical composition was observed between pristine and irradiated samples. Ti3AlC2 transforms to a cation-disordered Ti-Al-C system with the rocksalt structure after ion irradiation, in which Ti and Al atoms randomly occupy the cation sites and C atoms occupy the anion sites with occupancy of 0.5. The as-irradiated Ti3AlC2 exhibits a nano-twinned structure and the twin thickness is in the range between 2 and 10 nm. Most of defects are located at or near the twin boundaries, which may be attributed to the segregation of Al. Based on the structural variations of complex oxides induced by ion irradiation and stability of solid solution (Ti, Al)xC, it is assumed that Ti3AlC2 has a natural tendency to accommodate the defects through the chemical order to chemical disorder, low symmetry to high symmetry structural transformation. The formation mechanisms of the nano-twinned structure are related to the irradiation-induced antisite defect of cations, disorder of C atoms, and lattice structure characteristic of Ti3AlC2. Because the nano-twinned structure may significantly improve the irradiation tolerance of Ti(3)AIC(2) we study its stability under irradiation of Au ions at similar to 150 dpa. Only a moderate increase in the twin thickness is found and the nano-twinned structure is basically retained, which demonstrates the great stability of nano-twinned structure and makes Ti3AlC2 a very promising irradiation tolerant material for advanced nuclear systems
Large-Area and Visible-Light-Driven Heterojunctions of In<sub>2</sub>O<sub>3</sub>/Graphene Built for ppb-Level Formaldehyde Detection at Room Temperature
No full textAchieving convenient and accurate detection of indoor ppb-level formaldehyde is an urgent requirement to ensure a healthy working and living environment for people. Herein, ultrasmall In2O3 nanorods and supramolecularly functionalized reduced graphene oxide are selected as hybrid components of visible-light-driven (VLD) heterojunctions to fabricate ppb-level formaldehyde (HCHO) gas sensors (named InAG sensors). Under 405 nm visible light illumination, the sensor exhibits an outstanding response toward ppb-level HCHO at room temperature, including the ultralow practical limit of detection (pLOD) of 5 ppb, high response (Ra/Rg = 2.4, 500 ppb), relatively short response/recovery time (119 s/179 s, 500 ppb), high selectivity, and long-term stability. The ultrasensitive room temperature HCHO-sensing property is derived from visible-light-driven and large-area heterojunctions between ultrasmall In2O3 nanorods and supramolecularly functionalized graphene nanosheets. The performance of the actual detection toward HCHO is evaluated in a 3 m3 test chamber, confirming the practicability and reliability of the InAG sensor. This work provides an effective strategy for the development of low-power-consumption ppb-level gas sensors.Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.Bio-Electronic
Design of a novel ultra-wideband antenna applied for Life detecting
No full textThis paper introduces a novel antipodal tapered slot antenna based on the ultra-wideband characteristics of life detection radar. The antenna adopts edge comb-structure to improve the radiation field and uses the feeding of modified micro-strip line to slot line to solve the impedance matching problem of ultra-wideband antenna. The antenna is fabricated on F4B high-frequency board with the thickness of 1.5mm, and the size of the antenna is only 150mmx135mm. The measured results show that VSWR is lower than 2 in the operation frequency band of 900MHz to 5GHz, and the antenna gain is greater than 20dBi. The measured radiation patterns on E-plane and H-plane agree well with the simulated resultsCPCI-S(ISTP)[email protected]; [email protected]
A closed-form expansion approach for pricing discretely monitored variance swaps
No full textVariance swaps are among the most useful tools for derivatives trading and risk management. For pricing discretely monitored variance swaps under a general class of jump-diffusion models, we propose a closed-form expansion based on the length of monitoring interval. Our method relies on an iterative application of the Dynkin formula, which is usually called the operator method in financial econometrics. Numerical examples are given for demonstrating the efficiency of the method. (C) 2015 Elsevier B.V. All rights reserved.Guanghua School of Management at Peking University; Key Laboratory of Mathematical Economics and Quantitative Finance (Ministry of Education) at Peking University; Center for Statistical Sciences at Peking University; National Natural Science Foundation of China [11201009]; Department of Management Science and Engineering at Stanford UniversitySCI(E)[email protected]; [email protected]
Advances in the Fabrication, Mechanisms, and Applications of Monodisperse Droplets in Microfluidics
No full textMonodisperse microfluidic droplets, with precisely controlled size, high stability, and compartmentalization, have emerged as powerful tools in biomedicine, chemistry, and materials science. This review systematically summarizes key droplet generation methods, including T-junction, flow-focusing, and co-flow configurations, emphasizing how droplet size, frequency, and morphology are governed by channel geometry and operating parameters. Numerical modeling approaches–particularly Volume-of-Fluid (VOF), Level-Set (LS), and Phase-Field (PF) methods–are evaluated for their capabilities in capturing droplet formation dynamics and guiding device design, with VOF highlighted as the most reliable due to its mass-conservation properties. Applications of monodisperse droplets are further discussed in three major domains: biomedicine, chemical reactions, and materials fabrication. Overall, this review consolidates current advances in droplet fabrication, mechanisms, applications and outlines future directions to promote cross-disciplinary innovations
Ultrasmall fluorescent nanoparticles derived from roast duck: their physicochemical characteristics and interaction with human serum albumin
No full textFluorescent nanoparticles (FNPs) produced from roast meat have drawn widespread attention due to their potential hazards to human health.</p
A intervenção do professor e a competência comunicativa dos alunos de português como língua estrangeira/segunda : um estudo de caso na perspetiva do ensino dialógico
No full textNas últimas décadas, o conceito de competência comunicativa tem recebido
destaque no ensino-aprendizagem de línguas estrangeiras, onde o domínio das
habilidades de comunicação é considerado essencial nas interações. A intervenção do
professor, como parte importante da interação professor-aluno, desempenha um papel
insubstituível: não apenas melhora o input linguístico efetivo dos alunos, mas também
induz o output, construindo uma ponte de comunicação entre o ensino e a aprendizagem.
No entanto, a investigação atual sobre a intervenção do professor centra-se sobretudo
nas disciplinas de Matemática e Ciências, raramente estando relacionada com a
aprendizagem de línguas estrangeiras. Recentemente, é cada vez mais crucial
considerar o aluno como o construtor do seu próprio conhecimento, com o professor
assumindo um papel de facilitador, tal como na abordagem comunicativa no ensino de
línguas estrangeiras.
O ensino dialógico implica uma conversa contínua entre o professor e os alunos,
utilizando a comunicação eficaz para promover o ensino e a aprendizagem. Os
professores assumem a responsabilidade de criar diálogos que incentivem os alunos a
raciocinar, discutir e argumentar, em vez de apenas responderem passivamente. Desta
forma, o ensino dialógico pode oferecer uma base teórica para orientar a prática dos
professores em sala de aula e promover um ambiente de aprendizagem mais interativo
e reflexivo.
Tendo como objetivo entender o impacto de intervenção do professor em aulas de
português como língua estrangeira (PLE) na perspetiva do ensino dialógico, a autora
vai observar aulas de PLE em quatro níveis de proficiência (A2, B1, B2, C1). Através
das evidências concretas, como a natureza das perguntas feitas pelo professor ou a
forma como os professores respondem às contribuições dos alunos, pretende-se
identificar se as intervenções do professor promovem a participação ativa dos alunos e
incentivam a troca de ideias.
Através de grelhas de observação e anotações de observação da autora, as formas
mais comuns de intervenção são a explicação, a extensão e a sugestão. Os professores raramente usam a repetição, a complementação e o comentário para intervir e
acompanham os elogios com outras formas de intervenção. Não se observou o uso de
críticas nas aulas observadas.
A análise dos dados recolhidos revela que as formas de intervenção podem ser
classificadas de acordo com o seu foco, seja no conteúdo ou no processo. A explicação
e a extensão servem como um material de aprendizagem para oferecer aos alunos
oportunidade de consolidar e aperfeiçoar os conhecimentos já adquiridos. As
intervenções breves como a correção, o complemento e a sugestão envolvem menos
conhecimento específico de língua e focam mais em avançar os diálogos entre professor
e alunos. A repetição, o comentário e o elogio geralmente não abordam conhecimentos,
mas são cruciais para responder às contribuições dos alunos e manter o andamento da
aula. As diversas formas de intervenção do professor podem cultivar o pensamento
crítico, a capacidade de reflexão e a autonomia dos alunos.In recent decades, the concept of communicative competence has gained attention
in foreign language teaching and learning, where the mastery of communication skills
is considered essential in interactions. Teacher intervention, as an important part of
teacher-student interaction, plays an irreplaceable role: it not only improves students'
effective linguistic input, but also induces output, which builds a communication bridge
between teaching and learning. However, current research on teacher intervention
focuses mainly on mathematics and science and is rarely related to foreign language
learning. Recently, it has become increasingly crucial to consider the student as the
builder of their own knowledge, with the teacher taking on a facilitating role, as in the
communicative approach to foreign language teaching.
Dialogic teaching implies a continuous conversation between the teacher and the
students, using effective communication to promote teaching and learning. Teachers
take responsibility for creating dialogues that encourage students to reason, discuss and
argue, rather than just responding passively. In this way, dialogic teaching can offer a
theoretical basis to guide teachers' practice in the classroom, promoting a more
interactive and reflective learning environment.
With the aim of understanding the impact of teacher intervention in Portuguese as
foreign language classes from the perspective of dialogic teaching, the author will
observe PLE classes at four proficiency levels (A2, B1, B2, C1). Through concrete
evidence, such as the nature of the questions asked by the teacher or the way in which
teachers respond to student contributions, we try to identify whether the teacher's
interventions promote active student participation and encourage the exchange of ideas.
Through the author's observation grids and observation notes, explanation, extension
and suggestion are the most common forms. Teachers rarely use repetition,
complementation and commentary to intervene and accompany praise with other forms
of intervention. The use of criticism was not observed in the lessons observed.
Analysis of the data collected reveals that forms of intervention can be classified
according to their focus, whether on content or process. Explanation and extension serve as learning material to offer students the opportunity to consolidate the knowledge
they have already acquired. Short interventions such as correction, complement and
suggestion involve less specific language knowledge and focus more on advancing
dialogs between teacher and students. Repetition, comment and praise generally don't
address specific language knowledge, but are crucial for responding to students'
contributions and keeping the lesson on track. The various forms of teacher intervention
could cultivate students' critical thinking, ability to reflect and autonomy
Meta-Teacher For Face Anti-Spoofing
No full textAbstractFace anti-spoofing (FAS) secures face recognition from presentation attacks (PAs). Existing FAS methods usually supervise PA detectors with handcrafted binary or pixel-wise labels. However, handcrafted labels may are not the most adequate way to supervise PA detectors learning sufficient and intrinsic spoofing cues. Instead of using the handcrafted labels, we propose a novel Meta-Teacher FAS (MT-FAS) method to train a meta-teacher for supervising PA detectors more effectively. The meta-teacher is trained in a bi-level optimization manner to learn the ability to supervise the PA detectors learning rich spoofing cues. The bi-level optimization contains two key components: 1) a lower-level training in which the meta-teacher supervises the detector’s learning process on the training set; and 2) a higher-level training in which the meta-teacher’s teaching performance is optimized by minimizing the detector’s validation loss. Our meta-teacher differs significantly from existing teacher-student models because the meta-teacher is explicitly trained for better teaching the detector (student), whereas existing teachers are trained for outstanding accuracy neglecting teaching ability. Extensive experiments on five FAS benchmarks show that with the proposed MT-FAS, the trained meta-teacher 1) provides better-suited supervision than both handcrafted labels and existing teacher-student models; and 2) significantly improves the performances of PA detectors.Abstract
Face anti-spoofing (FAS) secures face recognition from presentation attacks (PAs). Existing FAS methods usually supervise PA detectors with handcrafted binary or pixel-wise labels. However, handcrafted labels may are not the most adequate way to supervise PA detectors learning sufficient and intrinsic spoofing cues. Instead of using the handcrafted labels, we propose a novel Meta-Teacher FAS (MT-FAS) method to train a meta-teacher for supervising PA detectors more effectively. The meta-teacher is trained in a bi-level optimization manner to learn the ability to supervise the PA detectors learning rich spoofing cues. The bi-level optimization contains two key components: 1) a lower-level training in which the meta-teacher supervises the detector’s learning process on the training set; and 2) a higher-level training in which the meta-teacher’s teaching performance is optimized by minimizing the detector’s validation loss. Our meta-teacher differs significantly from existing teacher-student models because the meta-teacher is explicitly trained for better teaching the detector (student), whereas existing teachers are trained for outstanding accuracy neglecting teaching ability. Extensive experiments on five FAS benchmarks show that with the proposed MT-FAS, the trained meta-teacher 1) provides better-suited supervision than both handcrafted labels and existing teacher-student models; and 2) significantly improves the performances of PA detectors
Closed-Form Expansions of Discretely Monitored Asian Options in Diffusion Models
No full textIn this paper we propose a closed-form asymptotic expansion approach to pricing discretely monitored Asian options in general one-dimensional diffusion models. Our expansion is a small-time expansion because the expansion parameter is selected to be the square root of the length of monitoring interval. This expansion method is distinguished from many other pricing-oriented expansion algorithms in the literature because of two appealing features. First, we illustrate that it is possible to explicitly calculate not only the first several expansion terms but also any general expansion term in a systematic way. Second, the convergence of the expansion is proved rigorously under some regularity conditions. Numerical experiments suggest that the closed-form expansion formula with only a few terms (e.g., four terms up to the third order) is accurate, fast, and easy to implement for a broad range of diffusion models, even including those violating the regularity conditions.Operations Research & Management ScienceMathematics, AppliedSCI(E)[email protected]; [email protected]; [email protected]
- …
