We study the problem of learning representations entities and relations in knowledge graphs for predicting missing links. The success such a task heavily relies on ability modeling inferring patterns (or between) relations. In this paper, we present new approach graph embedding called RotatE, which is able to model infer various relation including: symmetry/antisymmetry, inversion, composition. Specifically, RotatE defines each as rotation from source entity target complex vector space....

Recent work has explored the potential to adapt a pre-trained vision transformer (ViT) by updating only few parameters so as improve storage efficiency, called parameter-efficient transfer learning (PETL). Current PETL methods have shown that tuning 0.5% of parameters, ViT can be adapted downstream tasks with even better performance than full fine-tuning. In this paper, we aim further promote efficiency meet extreme constraint in real-world applications. To end, propose...

Navigation accuracy of an inertial navigation system can be significantly enhanced by rotating measurement unit with gimbals. Therefore, nonorthogonal angles gimbals, which are coupled into the error during rotation, should calibrated and compensated effectively. In this paper, relationship model is established. Then, calibration scheme observation equation gimbals rotation proposed. Proved a piecewise constant method, all parameters observable estimated extended Kalman filter. Experimental...

Sentence pair modeling is a crucial problem in the field of natural language processing. In this paper, we propose model to measure similarity sentence focusing on interaction information. We utilize word level matrix discover fine-grained alignment two sentences. It should be emphasized that each has different importance from perspective semantic composition, so exploit novel and efficient strategies explicitly calculate weight for word. Although proposed only use sequential LSTM without...

In inertial navigation system (INS)/doppler velocity log (DVL) integrated navigation, DVL commonly provides the 3-D of vehicle to filter. Theoretically, this INS/DVL fusion is loosely coupled approach, in which requires sufficient beam measurements (at least three) calculate velocity. However, cases that only has limited (fewer than measurements, possible underwater environment, can no longer work, and leaves INS work alone. Therefore, error will accumulate with time. contrast tightly...

Few-shot classification aims to recognize unseen classes with few labeled samples from each class. Many meta-learning models for few-shot elaborately design various task-shared inductive bias (meta-knowledge) solve such tasks, and achieve impressive performance. However, when there exists the domain shift between training tasks test obtained fails generalize across domains, which degrades performance of models. In this work, we aim improve robustness through task augmentation. Concretely,...

The pretrain-then-finetune paradigm has been widely adopted in computer vision. But as the size of Vision Transformer (ViT) grows exponentially, full finetuning becomes prohibitive view heavier storage overhead. Motivated by parameter-efficient transfer learning (PETL) on language transformers, recent studies attempt to insert lightweight adaptation modules (e.g., adapter layers or prompt tokens) pretrained ViT and only finetune these while weights are frozen. However, were originally...

Contrastive learning between different views of the data achieves outstanding success in field self-supervised representation and learned representations are useful broad downstream tasks. Since all supervision information for one view comes from other view, contrastive approximately obtains minimal sufficient which contains shared eliminates non-shared views. Considering diversity tasks, it cannot be guaranteed that task-relevant is Therefore, we assume ignored theoretically prove not...

Masked Image Modeling (MIM) achieves outstanding success in self-supervised representation learning. Unfortunately, MIM models typically have huge computational burden and slow learning process, which is an inevitable obstacle for their industrial applications. Although the lower layers play key role MIM, existing conduct reconstruction task only at top layer of encoder. The are not explicitly guided interaction among patches used calculating new activations. Considering requires non-trivial...

Gravity matching algorithm is a key technique of gravity aided navigation for underwater vehicles. The reliability traditional single point can be easily affected by environmental disturbance, which results in mismatching and decrease accuracy. Therefore, particle filter (PF)-based with sample vector proposed. correlation between adjacent points inertial system considered the order to solve problem. current sampling result rectified vectors composed selected point. amount determined field...

Time delay is a major problem in the acoustic communication technology. Considering such background, new dynamic model proposed for an improved error estimation algorithm this paper. And propagation equations are constructed inertial navigation system/Doppler velocity log integrated system. The time reconsidered with ultra-short baseline positioning system that uses multi-autonomous underwater vehicle (AUV) cooperative process master-slave mode. characteristics of delays considered and...

Autoregressive sequence models achieve state-of-the-art performance in domains like machine translation. However, due to the autoregressive factorization nature, these suffer from heavy latency during inference. Recently, non-autoregressive were proposed reduce inference time. assume that decoding process of each token is conditionally independent others. Such a generation sometimes makes output sentence inconsistent, and thus learned could only inferior accuracy compared their counterparts....

Dynamic deformation of a vehicle and relative position inertial navigation systems will cause large errors during information sharing. Such should be estimated compensated effectively, or it to output. Therefore, the model between angle dynamic lever arm is established verify that influence not coaxial but decussate. Then, estimation compensation methods for are proposed with real-time closed-loop correction lever-arm length. Simulation results demonstrate that, misalignment angle, method...

A novel adaptive finite-time control method is investigated for the gyros systems in presence of external disturbance. The fast nonsingular terminal sliding mode surface (NTSMS) designed which has all advantages NTSMS and general together. Based on law Lyapunov stability method, proposed theorems are given proven, we can obtain that synchronization errors converge to a small region including zero finite time. Furthermore, it be seen new algorithm guarantee chaotic have some excellent...

Gravity-aided inertial navigation is a leading issue in the application of autonomous underwater vehicle. An improved gravity matching algorithm, based on principle terrain contour (TERCOM) proposed. The algorithm applies shortest path to increase update frequency. In addition, positioning error can be limited due novel correlation analysis method. Compared with existing algorithms, TERCOM has better real-time performance, accuracy, and reduced calculation burden. reliability accuracy are...

Motivated by tracking applications with sensor networks under non-Gaussian noise and intermittent observations, this paper considers a maximum correntropy unscented Kalman filter (MCUKF). MCUKF is based on criterion (MCC) transformation (UT) which can deal both observations. The observations are described binary sequence satisfying some properties. MCC used to improves the robustness. Moreover, arrival probabilities (shot Gaussian mixture noise) given. performance of presented algorithm...

Multi-document summarization is of great value to many real world applications since it can help people get the main ideas within a short time.In this paper, we tackle problem extracting summary sentences from multi-document sets by applying sparse coding techniques and present novel framework challenging problem. Based on data reconstruction sentence denoising assumption, two-level representation model depict process summarization. Three requisite properties proposed form an ideal...

Navigation errors of inertial navigation system can be effectively restrained by rotating measurement units (IMU), enhancing accuracy in long-endurance navigation. However, nonorthogonal angles between axes are inevitable during manufacturing. Coupled into the error, lead to accuracy, especially attitude decline rotation. Therefore, calibration and compensation will enhance systemic reliability. In this study, we first established model, then estimated effect on sensitive IMU using proposed...

With the renaissance of neural network in recent years, relation classification has again become a research hotspot natural language processing, and leveraging parse trees is common effective method tackling this problem.In work, we offer new perspective on utilizing syntactic information dependency tree present position encoding convolutional (PECNN) based for classification.First, treebased features are proposed to encode relative positions words help enhance word representations.Then,...