Abstract Optimization of images with bad compositions has attracted increasing attention in recent years. Previous methods however seldomly consider image similarity when improving composition aesthetics. This may lead to significant content changes or bring large distortions, resulting an unpleasant user experience. In this paper, we present a new algorithm for aesthetics, while retaining faithful, as much possible, the original content. Our method computes improved using unified model...

The proton-induced degradation for InP/InGaAs heterojunction bipolar transistors (HBTs) is studied based on the calculation in energy range from displacement damage threshold to 100 MeV with analytical model of nonionizing loss (NIEL). experiments 3 and 10 proton radiation HBTs have been performed, which significant current gain predominant. increase base due increased recombination responsible noticeable degradation. Qualitative comparison NIEL results coefficients made good correlation found.

Presently, sentence-level researches are very significant in fields like natural language processing, information retrieval, machine translation etc. In this paper we present a practical task on sentence classification. The main purpose of work is to classify the abstract sentences scientific papers corpus built by ourselves into four categories- background, goal, method and result- which differ from each other common usage, so that can do further such as frequent pattern mining, extraction...

Fraud detection is an increasingly important and difficult task in the modern technological environment. Existing classification methods for fraud are based on call behavior attributes, such as duration frequency of callings. With evolution mode, self-attributes cannot detect fraudsters who camouflage themselves by behaving like normal accounts. So we propose Attention-based Graph Representation learning Model(AGRM), which takes into account node's its context information. In our model,...

Far-field speech recognition is a challenging task that conventionally uses signal processing beamforming to attack noise and interference problem. But the performance has been found usually limited due heavy reliance on environmental assumption. In this paper, we propose unified multichannel far-field system combines neural transformer-based Listen, Spell, Attend (LAS) system, which extends end-to-end further include enhancement. Such framework then jointly trained optimize final objective...

Triplet Markov fields (TMF) model is suitable for dealing with multi-class segmentation of non-stationary synthetic aperture radar (SAR) images. In this study, an algorithm using TMF edge location fast unsupervised SAR images proposed. The new can locate accurately reasonable computational cost. First the statistical characteristics multiplicative speckle noise in image, strength based on ratio exponentially weighted averages operator introduced into Turbopixels to obtain a superpixel graph...

By combining audio and visual modalities, the speech recognition systems achieve higher performance robustness. The fusion strategies to this point are mainly three types: feature level model decision fusion. In paper, we present a novel audio-visual framework, in which joint dimensionality reductional approach is used project features into more compact subspaces. With correlation preserving criteria, representations of projected will be able preserve conveyed original feattre space. At same...

Facing a large amount of entities appearing on the web, entity linking has recently become useful. It assigns an from resource to one name mention help users grasp meaning this mention. Unfortunately, many possible can be assigned Apparently, usually co-occurring mentions are related and considered together determine their best assignments. This approach is called collective often conducted based graph. However, traditional methods either consume much time due scale graph or obtain low...

This paper addresses the numerical and statistical stability in integration of inertial navigation system global satellite (GNSS) using Kalman filter (KF). Due to different units used horizontal vertical components geodetic (curvilinear) coordinates vicinity earth’s surface, i.e. radians former meters latter, covariance innovation vector KF, which should be inversed, can ill-conditioned, resulting severe instability. A simple rescaling method is proposed address this problem, specifically,...

With the rapid development of Internet Things (IoT), continuous emergence cyberattacks have brought great threat to security network. Intrusion Detection System (IDS) which can identify malicious network attacks has become a strong tool ensure security. Many deep learning-based approaches been used in IDS. However, most these researches ignore internal structural characteristics traffic, and cannot accurately learn key features traffic. Thus, they low accuracy classifying different kinds...

Traditional publish/subscribe systems suffer from either the limited expressiveness of events or complexity matching algorithms. Moreover, most them do not consider semantic information events. This study proposes a fuzzy with reasoning system, which employs ontology to model and subscriptions. The elaborate two main components in system: 1) Offline part domain defines concepts relationships between them, extra knowledge is supplement for ontology. 2) Online shows details published events,...

With the development and popularization of highspeed network technology, Intrusion Detection System need to process more data. In most cases, only a small part data which be processed is abnormal The will overwhelmed by normal A large number occupy resources IDS lead lot false alarms. These all bring troubles for Response administrators. this paper, we present method based on semi-supervised learning massive in high-speed network. We add some representative labeled unlabeled dataset, let...

Along with explosion of information, how to cluster large-scale documents has become more and important. This paper proposes a novel document clustering algorithm (CLCL) solve this problem. first constructs lexical chains from feature space reflect different topics which input contain, also can be separated into clusters by these chains. However, separation is too rough. So, idea self organizing mapping used optimize partition. For agglomerating semantic similarities one cluster, influences...