In the standard quantum theory, causal order of occurrence between events is prescribed, and must be definite. This has been maintained in all conventional scenarios operation for batteries. this study we take a step further to allow charging batteries an indefinite (ICO). We propose nonunitary dynamics-based protocol experimentally investigate using photonic switch. Our results demonstrate that both amount energy charged thermal efficiency can boosted simultaneously. Moreover, reveal...

Sequential incentive marketing is an important approach for online businesses to acquire customers, increase loyalty and boost sales. How effectively allocate the incentives so as maximize return (e.g., business objectives) under budget constraint, however, less studied in literature. This problem technically challenging due facts that 1) allocation strategy has be learned using historically logged data, which counterfactual nature, 2) both optimality feasibility (i.e., cost cannot exceed...

How to separate foreground from camera and background motions is a difficult problem for human action recognition in unconstrained environments. Although the existing interest point based methods have shown attractive results, they always come with high computational complexity lose their power cluttered field motion. In this paper, new spatio-temporal detector on flow vorticity proposed, which can not only suppress most of effects motion but also provide prominent points around key...

Description methods based on interest points and Bag-of-Words (BOW) model have gained remarkable success in human action recognition. Despite their popularity, the existing point detectors always come with high computational complexity lose power when camera is moving. Additionally, vector quantization procedure BOW ignores relationship between bases large reconstruction errors. In this paper, a spatio-temporal detector flow vorticity used, which can not only suppress most effects of motion...

Targets detection and segmentation in a synthetic aperture radar (SAR) image is vital step for its interpretation. It quite challenging most conventional methods due to complex background the speckle. Furthermore, sizes of targets scene are variable. Inspired by success neural networks computer vision, In this paper, we propose 3D dilated multi-scale U-shape convolutional network (3DDM-UNet). proposed method, first build block via multiscale stationary wavelet transform exploit structural...

Modelling the user's multiple behaviors is an essential part of modern e-commerce, whose widely adopted application to jointly optimize click-through rate (CTR) and conversion (CVR) predictions. Most existing methods overlook effect two key characteristics behaviors: for each item list, (i) contextual dependence refers that on any are not purely determinated by itself but also influenced previous (e.g., clicks, purchases) other items in same sequence; (ii) time scales means users likely...

In this paper, we present a common framework of concept-based video retrieval and propose several methods to improve the performance system. 12 kinds features, including color, texture, shape local features are examined, modified HOG which is defined on image edges reduce its computational complexity. The concept cooccurrence matrix assistant (B&W detection, audio detection motion detection) suggested enhance Extensive experiments TRECVID 2010 show effectiveness our proposed methods.

With the rise of smart devices, there is an escalating need for lightweight methods in human pose estimation (HPE). Although existing 2D HPE techniques have demonstrated impressive performance on public datasets, they still suffer from high model complexity and latency issues practical applications. To address challenge, this paper proposes a novel approach HPE. By utilizing shuffle blocks instead traditional ResNet, we significantly reduce size computational requirements. Moreover, our...

Manually annotating datasets for training deep models is very labor-intensive and time-consuming. To overcome such inferiority, directly leveraging web images to conduct data becomes a natural choice. Nevertheless, the presence of label noise in usually degrades model performance. Existing methods combating are typically designed tested on synthetic noisy datasets. However, they tend fail achieve satisfying results real-world this end, we propose method named GRIP alleviate problem both...

In this paper, Sparse Coding with Non-negative and Locality constraints (SCNL) is proposed to generate discriminative feature descriptions for human action recognition. The non-negative constraint ensures that every data sample in the convex hull of its neighbors. locality makes a only represented by related neighbor atoms. sparsity confines dictionary atoms involved representation as fewer possible. SCNL model can better capture global subspace structures than classical sparse coding, are...

Human action recognition is a challenge problem in computer vision. In this paper, we propose an improved approach using kinematic features for recognition. approach, find the area that relates to by simple method, and select eight discriminative derived from optical flow field describe dynamics of field. The covariance matrix feature vectors used fuse serve as descriptor. Multi-class SVM classifiers are then employed classification. Experiments carried out on public datasets. We obtain rate...

Operations performing on quantum batteries are extended to scenarios where we no longer force the existence of definite causal order occurrence between distinct processes. In contrast standard theories, so called indefinite is found have capability accomplishing tasks that not possible without it. Specifically, show how this novel class resource comes into play in by first, combining two static unitary chargers a coherently superposed one fully charge an empty battery even if presence...

With the prevalence of live broadcast business nowadays, a new type recommendation service, called recommendation, is widely used in many mobile e-commerce Apps. Different from classical item to automatically recommend user anchors instead items considering interactions among triple-objects (i.e., users, anchors, items) rather than binary between users and items. Existing methods based on objects, ranging early matrix factorization recently emerged deep learning, obtain objects' embeddings...

A discriminative multi-modality non-negative sparse (DMNS) graph model is proposed in this paper. In the model, features each modality are first projected into Mahalanobis space by a transformation learned for modality, then constructed with shared coefficients across modalities. Both labeled and unlabeled data can be introduced graph, label propagation performed to predict labels of samples. Extensive experiments over two benchmark datasets demonstrate advantages DMNS-graph method...

Manually annotating datasets for training deep models is very labor-intensive and time-consuming. To overcome such inferiority, directly leveraging web images to conduct data becomes a natural choice. Nevertheless, the presence of label noise in usually degrades model performance.Existing methods combating are typically designed tested on synthetic noisy datasets. However, they tend fail achieve satisfying results real-world this end, we propose method named GRIP alleviate problem both...

AdaBoost based training method has become a state-of-the-art boosting approach in face detection system. In this paper, compared to the naive method, Forward Feature Selection (FFS) is used feature selection reduce time by about 50 100 times without loss of performance. Furthermore, hierarchical spaces (both local and global) construct detector cascade on FFS are adopted, which still have good discrimination later stage process. Experimental results show that our can achieve higher...

In all existing quantum walk models, the assumption about a pre-existing fixed background causal structure is always made and has been taken for granted. Nevertheless, in this work we will get rid of tacit especially by introducing indefinite order ways coin tossing investigate modern scenario. We find that an ideal-shape fast-spreading uniform distribution can be prepared with our new model. First show how always-symmetrical instantaneous appears walk, which then paves way deriving...

With the prevalence of live broadcast business nowadays, a new type recommendation service, called recommendation, is widely used in many mobile e-commerce Apps. Different from classical item to automatically recommend user anchors instead items considering interactions among triple-objects (i.e., users, anchors, items) rather than binary between users and items. Existing methods based on objects, ranging early matrix factorization recently emerged deep learning, obtain objects' embeddings...

Modelling the user's multiple behaviors is an essential part of modern e-commerce, whose widely adopted application to jointly optimize click-through rate (CTR) and conversion (CVR) predictions. Most existing methods overlook effect two key characteristics behaviors: for each item list, (i) contextual dependence refers that on any are not purely determinated by itself but also influenced previous (e.g., clicks, purchases) other items in same sequence; (ii) time scales means users likely...