Image aesthetics assessment has emerged as a hot topic in recent years due to its potential numerous high-level vision applications. In this paper, distinguished from existing studies relying on single label, we propose quantifying image by distribution over multiple quality levels. The distribution-based representation characterizes the disagreement among users' aesthetic preferences regarding same image, and is also compatible with traditional task of label prediction. Our framework...

Knowledge tracing aims to trace students’ evolving knowledge states by predicting their future performance on concept-related exercises. Recently, some graph-based models have been developed incorporate the relationships between exercises improve tracing, but only a single type of relationship information is generally explored. In this article, we present novel Dual Graph Ensemble learning method for Tracing (DGEKT), which establishes dual graph structure interactions capture heterogeneous...

Hashing has recently received great attention in cross-modal retrieval. Cross-modal retrieval aims at retrieving information across heterogeneous modalities (e.g., texts vs. images). hashing compresses high-dimensional data into compact binary codes with similarity preserving, which provides efficiency and facility both storage. In this study, we propose a novel fast discrete (FDCH) method regressing from semantic labels to take advantage of supervised improve performance. contrast existing...

In the past few years, micro-videos have become dominant trend in social media era. Meanwhile, as number of microvideos increases, users are frequently overwhelmed by their uninterested ones. Despite success existing recommendation systems developed for various communities, they cannot be applied to routing micro-videos, since micro-video platforms unique characteristics: diverse and dynamic interest, multilevel well true negative samples. To address these problems, we present a temporal...

Content-based near-duplicate video detection (NDVD) is essential for effective search and retrieval, robust fingerprinting a good solution NDVD. Most existing methods use single feature or concatenate different features to generate fingerprints, show performance under single-mode modifications such as noise addition blurring. However, when they suffer combined modifications, the degraded certain extent because cannot characterize content completely. By contrast, assistance consensus among...

Multi-view hashing can well support large-scale near-duplicate video retrieval, due to its desirable advantages of mutual reinforcement multiple features, low storage cost, and fast retrieval speed. However, there are still two limitations that impede performance. First, existing methods only consider local structures in features. They ignore the global structure is important for cannot fully exploit dependence complementarity Second, works always learn functions bit by bit, which...

Recently, the valuation of variable annuity products has become a hot topic in actuarial science. In this paper, we use Fourier cosine series expansion (COS) method to value guaranteed minimum death benefit (GMDB) products. We first express GMDB by discounted density function approach, then COS approximate Equations. When distribution time-until-death random is approximated combination exponential distributions and price fund modeled an Lévy process, explicit equations for coefficients are...

As mobile device penetration increases, it has become pervasive for images to be associated with locations in the form of geotags. Geotags bridge gap between physical world and cyberspace, giving rise new opportunities extract further insights into user preferences behaviors. In this article, we aim exploit geotagged photos from online photo-sharing sites purpose personalized Point-of-Interest (POI) recommendation. Owing fact that most users have only very limited travel experiences, data...

Local feature points have been widely employed in robust image fingerprinting. One of their intrinsic advantages is invariance under geometric transforms. However, robustness against certain attacks that modify the positions points, such as additive noising and blurring, limited. In addition, local-feature-point-based approaches ignore distribution points. this paper, we harness point relationships, including local structures global relevance, to overcome these limitations. relationship...

In this paper, we assume that the reserve level of an insurance company can only be observed at discrete time points, then a new risk model is proposed by introducing periodic capital injection strategy and barrier dividend into classical model. We derive equations boundary conditions satisfied Gerber-Shiu function, expected discounted function assuming observation interval claim amount are exponentially distributed, respectively. Numerical examples also given to further analyze influence...

In this work, a diagnostic scheme for lower atmospheric ducts was established based on the Weather Research and Forecasting (WRF) model. More specifically, 10-year simulation test conducted China seas to investigate spatio-temporal characteristics of phenomenon. Compared with sounding data, long-term simulations showed high temporal correlation root mean square error modified refractivity remained between 4 M 7 M. Based simulations, significant regional differences in occurrence probability...

Unsupervised domain adaptation (UDA) aims to adapt models learned from a well-annotated source target domain, where only unlabeled samples are available. To this end, adversarial training is widely used in conventional UDA methods reduce the discrepancy between and domains. Recently, prompt tuning has emerged as an efficient way large pre-trained vision-language like CLIP variety of downstream tasks. In paper, we present novel method named Adversarial DuAl Prompt Tuning (ADAPT) for UDA,...

Test-time prompt tuning (TPT) aims to adjust the vision-language models (e.g., CLIP) with learnable prompts during inference phase. However, previous works overlooked that pre-trained as a service (MaaS) have become noticeable trend due their commercial usage and potential risk of misuse. In context MaaS, users can only design in inputs query black-box through APIs, rendering paradigm utilizing gradient for is infeasible. this paper, we propose test-time (B²TPT), novel framework addresses...

A mobile ultrasonic stratified flow velocity measurement device, which utilizes a pair of transducers, was characterized by its low power consumption and the ability to measure multi-layer velocities in channels, offering advantages water applications agricultural irrigation areas. The study combined experimental research with computational modeling investigate impact propagation time downstream upstream flows effect position on prediction. Traditional time-difference methods were found be...

Relation classification is a crucial ingredient in numerous information-extraction systems and has attracted great deal of attention recent years. Traditional approaches largely rely on feature engineering suffer from the limitations domain adaption error propagation. To overcome above-mentioned problems, many deep neural network-based methods have been proposed; however, these cannot effectively locate utilize relation trigger features. features make full use them, we propose novel...