Relevance feedback is often a critical component when designing image databases. With these databases it difficult to specify queries directly and explicitly. interactively determinines user's desired output or query concept by asking the user whether certain proposed images are relevant not. For relevance algorithm be effective, must grasp accurately quickly, while also only label small number of images. We propose use support vector machine active learning for conducting effective...

Spectral clustering algorithms have been shown to be more effective in finding clusters than some traditional algorithms, such as k-means. However, spectral suffers from a scalability problem both memory use and computational time when the size of data set is large. To perform on large sets, we investigate two representative ways approximating dense similarity matrix. We compare one approach by sparsifying matrix with another Nyström method. then pick strategy via retaining nearest neighbors...

Relevance feedback is often a critical component when designing image databases. With these databases it difficult to specify queries directly and explicitly. interactively determinines user's desired output or query concept by asking the user whether certain proposed images are relevant not. For relevance algorithm be effective, must grasp accurately quickly, while also only label small number of images. We propose use support vector machine active learning for conducting effective...

We propose a content-based soft annotation (CBSA) procedure for providing images with semantical labels. The starts labeling small set of training images, each one single label (e.g., forest, animal, or sky). An ensemble binary classifiers is then trained predicting membership images. applied to individual image give the multiple labels, and associated factor. To select base binary-classifier CBSA, we experiment two learning methods, support vector machines (SVMs) Bayes point (BPMs), compare...

Frequent itemset mining (FIM) is a useful tool for discovering frequently co-occurrent items. Since its inception, number of significant FIM algorithms have been developed to speed up performance. Unfortunately, when the dataset size huge, both memory use and computational cost can still be prohibitively expensive. In this work, we propose parallelize FP-Growth algorithm (we call our parallel PFP) on distributed machines. PFP partitions computation in such way that each machine executes an...

With the revival of neural networks, many studies try to adapt powerful sequential models, ıe Recurrent Neural Networks (RNN), recommendation. RNN-based networks encode historical interaction records into a hidden state vector. Although vector is able dependency, it still has limited representation power in capturing complicated user preference. It difficult capture fine-grained preference from sequence. Furthermore, latent usually hard understand and explain. To address these issues, this...

An imbalanced training data set can pose serious problems for many real-world mining tasks that employ SVMs to conduct supervised learning. In this paper, we propose a kernel-boundary-alignment algorithm, which considers THE imbalance as prior information augment improve class-prediction accuracy. Using simple example, first show suffer from high incidences of false negatives when the instances target class are heavily outnumbered by nontarget class. The remedy is adjust boundary modifying...

In recent years, the boundaries between e-commerce and social networking have become increasingly blurred. Many Web sites support mechanism of login where users can sign on using their network identities such as Facebook or Twitter accounts. Users also post newly purchased products microblogs with links to product pages. this paper, we propose a novel solution for cross-site cold-start recommendation, which aims recommend from at in "cold-start" situations, problem has rarely been explored...

To answer user queries efficiently, a stream management system must handle continuous, high-volume, possibly noisy, and time-varying data streams. One major research area in seeks to allocate resources (such as network bandwidth memory) query plans, either minimize resource usage under precision requirement, or maximize of results constraints. date, many solutions have been proposed; however, most are ad hoc with hard-coded heuristics generate plans. In contrast, we perceive fundamentally...

Considerable research has been devoted to utilizing multimodal features for better understanding multimedia data. However, two core issues have not yet adequately addressed. First, given a set of extracted from multiple media sources (e.g., the visual, audio, and caption track videos), how do we determine best modalities? Second, once modalities identified, fuse them map semantics? In this paper, propose two-step approach. The first step finds statistically independent raw features. second...

A distributed data-stream architecture finds application in sensor networks for monitoring environment and activities. In such a network, large numbers of sensors deliver continuous data to central server. The rate at which the is sampled each affects communication resource computational load this paper, we propose novel adaptive sampling technique where adapts streaming-data characteristics. Our approach employs Kalman-Filter (KF)-based estimation wherein can use KF error adaptively adjust...

We propose using one-class, two-class, and multiclass SVMs to annotate images for supporting keyword retrieval of images. Providing automatic annotation requires an accurate mapping images' low-level perceptual features (e.g., color texture) some high-level semantic labels landscape, architecture, animals). Much work has been performed in this area; however, there is a lack ability assess the quality annotation. In paper, we confidence-based dynamic ensemble (CDE), which employs three-level...

Previous methods of distributed Gibbs sampling for LDA run into either memory or communication bottlenecks. To improve scalability, we propose four strategies: data placement , pipeline processing word bundling and priority-based scheduling . Experiments show that our strategies significantly reduce the unparallelizable bottleneck achieve good load balancing, hence scalability LDA.

Two characteristics of location-based services are mobile trajectories and the ability to facilitate social networking. The recording trajectory data contributes valuable resources towards understanding users’ geographical movement behaviors. Social networking is possible when users able quickly connect anyone nearby. A network with location based known as (LBSN). As shown in Cho et al. [2013], locations that frequently visited by socially related persons tend be correlated, which indicates...

Users of social networking services can connect with each other by forming communities for online interaction. Yet as the number hosted such websites grows over time, users have even greater need effective community recommendations in order to meet more users. In this paper, we investigate two algorithms from very different domains and evaluate their effectiveness personalized recommendation. First is association rule mining (ARM), which discovers associations between sets that are shared...

Layered growth of molybdenum disulphide (MoS 2 ) was successfully achieved by pulsed laser deposition (PLD) method on c ‐plane sapphire substrate. Growth monolayer to a few MoS , dependent the number excimer in PLD is demonstrated, indicating promising controllability layer growth. Among samples with various pulse deposition, frequency difference (A 1g –E 1 2g Raman analysis 70 sample estimated as 20.11 cm –1 suggesting obtained. Two‐dimensional (2D) confirmed streaky reflection high energy...

Monolithic 3D integrated circuits using transition metal dichalcogenide materials and low-temperature processing are reported. A variety of digital analog implemented on two sequentially layers devices. Inverter circuit operation at an ultralow supply voltage 150 mV is achieved, paving the way to high-density, ultralow-voltage, ultralow-power applications.

There are two major challenges to overcome when developing a classifier perform automatic disease diagnosis. First, the amount of labeled medical data is typically very limited, and cannot be effectively trained attain high disease-detection accuracy. Second, domain knowledge required identify representative features in for detecting target disease. Most computer scientists statisticians do not have such knowledge. In this work, we show that employing transfer learning can remedy both...

60 nm InAs high-electron-mobility transistors (HEMTs) with a thin channel, InAlAs barrier layer, and very high gate stem structure have been fabricated characterized. The thickness of the as well that was reduced to 5 nm. A height 250 Pt-buried used in device configuration reduce parasitics. DC transconductance 2114 mS/mm current-gain cutoff frequency (fT) 710 GHz were achieved at VDS=0.5 V.

In this letter, a gate recessed normally OFF AlGaN/GaN MIS-HEMT with low threshold voltage hysteresis using Al <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> O xmlns:xlink="http://www.w3.org/1999/xlink">3</sub> /AlN stack insulator is presented. The trapping effect of /GaN interface was effectively reduced the insertion 2-nm AlN thin interfacial passivation layer grown by plasma enhanced atomic deposition. device exhibits +1.5 V,...