As an important and challenging problem in the multimedia area, multi-modal data understanding aims to explore intrinsic semantic information across different modalities a collaborative manner. To address this problem, possible solution is effectively adaptively capture common cross-modal by modeling inherent correlations between latent topics from modalities. Motivated task, we propose supervised mutual topic reinforce (M$^3$R) approach, which seeks build joint probabilistic graphical model...

In multimedia information retrieval, most classic approaches tend to represent different modalities of media in the same feature space. With click data collected from users' searching behavior, existing take either one-to-one paired (text-image pairs) or ranking examples (text-query-image and/or image-query-text lists) as training examples, which do not make full use data, particularly implicit connections among objects. this paper, we treat a large graph, vertices are images/text queries...

A nonreciprocal magnon laser is proposed in a compound cavity optomagnonical system consisting of an yttrium iron garnet sphere coupled to spinning resonator. On the basis magnon-induced Brillouin scattering process making it possible achieve lasing action, Fizeau light-dragging effect caused by resonator further results significant modifications gain and threshold power for different driving directions, then realized. Especially, this highly tunable speed direction. Our work provides...

We investigate magnon statistics in a qubit-magnon hybrid quantum system which an effective appreciable coupling can be realized by exchanging virtual cavity photons. A conventional blockade and two types of unconventional blockades are proposed, respectively, based on three different physical mechanisms. verify theoretically that occur strong, weak, moderate regimes. It is interesting asymmetry structure for anti-bunching observed the case strength, especially where interference...

Electronic medical records (EMRs) play an important role in data mining and sequential learning. In this article, we propose to use a neural network with dynamic content-based memories predict future medications, given EMRs. The local–global memory contains two layers of memories: the local global memory. Particularly, our method learns hidden knowledge within EMRs by locally remembering individual patterns patient (via memory) globally group evidence disease memory). addition, show how...

The query logs from an on-line map system provide rich cues to understand the behaviors of human crowds. With growing ability collecting large scale logs, suggestion has been a topic recent interest. In general, aims at recommending list relevant queries w.r.t. users' inputs via appropriate learning crowds' logs. this paper, we are particularly interested in suggestions (e.g., predictions location-related queries) and propose novel model <italic xmlns:mml="http://www.w3.org/1998/Math/MathML"...

In this paper, we propose an approach for generating rich fine-grained textual descriptions of images. particular, use LSTM-in-LSTM (long short-term memory) architecture, which consists inner LSTM and outer LSTM. The effectively encodes the long-range implicit contextual interaction between visual cues (i.e., spatiallyconcurrent objects), while generally captures explicit multi-modal relationship sentences images correspondence images). This architecture is capable producing a long...

For a number of important problems, isolated semantic representations individual syntactic words or visual objects do not suffice, but instead compositional representation is required; for example, literal phrase set spatially concurrent objects. In this paper, we aim to harness the existing image-sentence databases exploit nature data multimodal deep embedding. particular, propose an approach called hierarchical-alike (bottom-up two layers) grounded semantics (hiMoCS) learning. The proposed...

Multi-criteria QoS routing has been regarded by researchers as a vital mechanism to support multimedia communications. Unfortunately, finding qualified path meeting the multiple criteria is indeed very difficult problem solve. Multi-path-constrained problems are considered be NP complete, and most existing algorithms based on maintaining global network state at every node. This paper proposes multi-constrained flooding-based with minimum computation implementation complexity which relies...

The objective of this paper is to propose a hierarchical topic evolution model ( HTEM ) that can organize time‐varying topics in hierarchy and discover their evolutions with multiple timescales. In the proposed , near root are more abstract also evolve longer timescales than those leaves. To achieve goal, distance‐dependent C hinese restaurant process (dd CRP extended new nested able simultaneously dependencies among data relationship between clusters. based on for time‐stamped documents,...

We consider the problem of embedding entities and relations knowledge bases into low-dimensional continuous vector spaces (distributed representations). Unlike most existing approaches, which are primarily efficient for modelling pairwise between entities, we attempt to explicitly model both long-range interactions by interpreting them as linear operators on embeddings entities. Therefore, in this paper introduces Path-Ranking capture graph at same time preserve graph; call it 'structured...

A novel efficient scheme for deterministic secure quantum communication with cluster state is proposed. By utilizing Bell-basis- and [Formula: see text]-basis-measurements, the two legitimate users can directly transmit secret message after exchanging some additional classical information. It has a high capacity as each carry bits of information, intrinsic efficiency because almost all instances are useful. Since present based on which robust against decoherence, it easily processed by...

We propose a new protocol for deterministic secure quantum communication with four-qubit W state. In this protocol, by using four local two-particle unitary operations and Bell basis measurements, the two legitimate users can directly transmit secret messages after exchanging some additional classical information. It will be shown that our has high capacity as each state carry bits of information, intrinsic efficiency because almost all instances are useful. Furthermore, is feasible present...

We propose two schemes for quantum secure direct communication (QSDC) and deterministic (DSQC) over collective dephasing noisy channel. In our schemes, four special two-qubit states are used as the Since these unchanged through channel, effect of channel noise can be perfectly overcome. Simultaneously, security against some usual attacks ensured by utilizing various checking procedures. Furthermore, feasible with present-day technique.

In the context of quantum mechanics we employ technique integration within an ordered product operators to recast classical wavelet transform a squeezing-displacing between mother vector and state be transformed. this way propose wavelet-transform spectrum for optical states. For some typical states obtain numerical results which imply that can used recognize variety states, inverse has possibility play role in engineering.

We propose a scheme for sharing an arbitrary unknown two-qubit state among three parties by using four-qubit cluster-class and Bell as quantum channel. With controlled phase gate (QCPG) operation local unitary operation, any one of the two agents has access to reconstruct original if he/she collaborates with other one, whilst individual agent obtains no information. As all resource can be used carry useful information, intrinsic efficiency qubits approaches maximal value. Moreover, present...

We construct a new type of photon-subtracted squeezed coherent state (PSSCS) based on (SCS) [Phys. Lett. A 220 (1996) 81]. Some the statistical properties PSSCS, such as Mandel's Q parameter and photon-number distribution, are investigated corresponding non-classicality is discussed.

In the context of quantum mechanics classical wavelet transform a function f with mother can be recast into matrix element squeezing-displacing operator U(,s) as 〈|U(,s) |f〉. The technique integral within an ordered product operators is used to support this theory. Based on this, transforms are done for even- and odd-binomial states, corresponding numerical calculation leads spectrum transform, which helpful recognizing difference between odd-states.