Abstract Objective. Parkinson’s disease (PD) is one of the most common neurodegenerative diseases, and early diagnosis crucial to delay progression. The PD has always been a difficult clinical problem due lack reliable biomarkers. Electroencephalogram (EEG) detection method, studies have attempted discover EEG spectrum characteristics PD, but reported conclusions are not uniform heterogeneity patients. There an urgent need for more advanced algorithm extract from satisfy personalized...

5-Spiro-cyclopentene-rhodanine derivatives have been synthesized through two different diastereoselective phosphine-catalyzed cascade reactions in excellent diastereoselectivities and good to yields.

The AntClass algorithm is a new applying ant colony clustering to cluster analysis, and the result satisfying. To attack slow speed of algorithm, named DBAntCluster proposed. Firstly, high density clusters are got in dataset by using DBSCAN then these scattered grid board as special kind data object with other single objects dataset. In ants can avoid many unnecessary movements attribute distribution well, greatly accelerated. This improvement validated our experiments.

As the radio communications technology widely used, wireless location plays a more important role in maintaining order of air waves. However concretely effective symbol calibration method with regard to Chinese DTMB signal different frame mode is quite under research due multiple structure signal. In this paper, we propose Time Difference Arrival(TDOA)- based passive scheme using least square principle. Utilizing large number anchor nodes monitoring network, novel algorithm formulated solve...

Geo-textual objects with both geographical location and textual description are gaining in prevalence. Over the past decades, substantial research has been conducted on spatial keyword queries, which integrate into keyword-based querying of geo-textual content. However, existing proposals mostly focus efficiency for processing little effort was made to address effectiveness perspectives. In this work, using two datasets ground truth query results, we evaluate standard queries. Our evaluation...

In this paper, we present a new method to improve the performance of current bag-of-word based image classification process. After feature extraction, introduce pair wise matching scheme select discriminative features. Only label information from raining-sets is used update weights via an iterative processing. The selected features correspond foreground content images thus highlight high level category knowledge images. "Visual words" are constructed on these Our can be as refinement step...

Micro-video recommendation is important for micro-video social platform, which provides its users with micro-videos they may interested in. In this paper, we propose a new variant of Transformer, alleviating the drawbacks Recurrent Neural Networks (RNN) tend to compress all history records in fixed hidden representation, model users' sequential behavior next recommendation. Firstly, employ self-attention capture micro-video's multi-modal features different importance. Secondly, make use...

When a user tend to publish micro-video, hashtag recommendation aims suggest hashtags that can reflect the theme or contents of and meet tagging preference as well. In this paper, we show how profile historical combined with micro-video representations be used perform recommendation. Specifically, User-guided Hierarchical Multi-head Attention Network (UHMAN) is proposed attend both image-level video-level micro-videos side information. We evaluate model on dataset collected from sharing...

Uncertainty sampling is a prevalent active learning algorithm that queries sequentially the annotations of data samples which current prediction model uncertain about. However, usage uncertainty has been largely heuristic: (i) There no consensus on proper definition "uncertainty" for specific task under loss; (ii) theoretical guarantee prescribes standard protocol to implement algorithm, example, how handle arrived annotated framework optimization algorithms such as stochastic gradient...

Based on the result of domestic and international highway risks subjects behavioral risks, starting from factor entities ability, this paper identified main abilities management entities. It explained concept ability analyzed influence which every kind affect four behavior (Government behavior, Internal Highway enterprises, Competitor customer behavior). used notional model to elaborate formation mechanism subjects.

Many universities offer artificial intelligence courses. Not only the content of course is abundant and intricate, but also much complicated calculation contained in course. The study found that it difficult to learn this by traditional teaching methods depend on blackboard textbooks. So, paper presents a design MOOC for as follows. Firstly constructs knowledge system Secondly, gives different content. Thirdly, construction MOOC, which includes video, forums, exercises testing. uses...

In this work, we propose a Unified framework of Sequential Search and Recommendation (UnifiedSSR) for joint learning user behavior history in both search recommendation scenarios. Specifically, consider user-interacted products the scenario, user-issued queries scenario as three distinct types behaviors. We dual-branch network to encode pair interacted product issued query parallel. This allows cross-scenario modeling by deactivating branch scenario. Through parameter sharing between dual...

With the proliferation of spatio-textual data, Top-k KNN spatial keyword queries (TkQs), which return a list objects based on ranking function that evaluates both and textual relevance, have found many real-life applications. Existing geo-textual indexes for TkQs use traditional retrieval models like BM25 to compute text relevance usually exploit simple linear but its effectiveness is limited. To improve effectiveness, several deep learning recently been proposed, they suffer severe...

In this paper, we consider the supervised pretrained transformer for a class of sequential decision-making problems. The considered problems is subset general formulation reinforcement learning in that there no transition probability matrix, and covers bandits, dynamic pricing, newsvendor as special cases. Such structure enables use optimal actions/decisions pretraining phase, usage also provides new insights training generalization transformer. We first note model can be viewed performative...

We introduce MVGenMaster, a multi-view diffusion model enhanced with 3D priors to address versatile Novel View Synthesis (NVS) tasks. MVGenMaster leverages that are warped using metric depth and camera poses, significantly enhancing both generalization consistency in NVS. Our features simple yet effective pipeline can generate up 100 novel views conditioned on variable reference poses single forward process. Additionally, we have developed comprehensive large-scale image dataset called...

When doing clustering analysis it always needs a validity index to evaluate if the present scheme can reflect real natural structure of dataset. The clusters founded by algorithm be arbitrary shape, but exiting indices only assess convex clusters. To solve this problem new CompSepa is proposed in paper, which cluster including both non-convex and clusters, computed minimum-cost spanning tree (MST) objects Experiments show that correctly effectively.

In this work, we propose a Unified framework of Sequential Search and Recommendation (UnifiedSSR) for joint learning user behavior history in both search recommendation scenarios. Specifically, consider user-interacted products the scenario, as well user-issued queries scenario three distinct types behaviors. We dual-branch network to encode pair interacted product issued query parallel. This allows cross-scenario modeling by deactivating branch scenario. Through parameter sharing between...

DBSCAN is a classic density based algorithm and it clusters the data set according to user input parameters. This work investigates how inherit mining results of last time when parameters change. A new incremental clustering IPC-DBSCAN proposed, which gets same result as that rerunning yet high efficiency achieved. Theoretical analysis experimental show proposed method reduces search space greatly has novel efficiency. By interaction, most satisfying quickly especially suits large volume set.

Learning with rejection has been a prototypical model for studying the human-AI interaction on prediction tasks. Upon arrival of sample instance, first uses rejector to decide whether accept and use AI predictor make or reject defer humans. such changes structure original loss function often results in undesirable non-convexity inconsistency issues. For classification problem, several works develop consistent surrogate losses joint learning rejector, while there have fewer regression...