Deep learning-based methods have demonstrated significant breakthroughs in the application of hyperspectral image (HSI) classification. However, some challenging issues still exist, such as overfitting problem caused by limitation training size with high-dimensional feature and efficiency spectral–spatial (SS) exploitation. Therefore, to efficiently model relative position samples within generative adversarial network (GAN) setting, we proposed a dual-channel SS fusion capsule (DcCapsGAN)...

Deep learning (DL) has become a hot topic in the research field of hyperspectral image (HSI) classification. However, with increasing depth and size deep methods, its application mobile embedded vision applications brought great challenges. In this article, we address network architecture search (NAS)-guided lightweight spectral–spatial attention feature fusion (LMAFN) for HSI The overall proposed is guided by several conclusions NAS, which achieves fewer parameters lower computation cost...

The Kennedy-Tasaki (KT) transformation was used to construct the gapped symmetry protected topological (SPT) phase from breaking with open boundary condition, and generalized in our proceeding work [L. Li et al. arXiv:2301.07899] on a ring by sacrificing unitarity, should be understood as non-invertible duality transformation. In this work, we further apply KT systematically gapless phases. This construction reproduces known examples of (intrinsically) SPT where non-trivial features come...

Few-Shot Class-Incremental Learning (FSCIL) aims to continuously learn new classes from a limited set of training samples without forgetting knowledge previously learned classes. Conventional FSCIL methods typically build robust feature extractor during the base session with abundant and subsequently freeze this extractor, only fine-tuning classifier in subsequent incremental phases. However, current strategies primarily focus on preventing catastrophic forgetting, considering relationship...

We study quantum many-body systems in the presence of an exotic antiunitary translation or inversion symmetry involving time reversal. Based on a symmetry-twisting method and spectrum robustness,we propose that half-integer spin chain which respects any these two crystalline symmetries addition to discrete $\mathbb{Z}_2\times\mathbb{Z}_2$ global spin-rotation must either be gapless possess degenerate ground states. This explains gaplessness class chiral models not indicated by...

It is still challenging to effectively detect ship objects in optical remote-sensing images with complex backgrounds. Many current CNN-based one-stage and two-stage detection methods usually first predefine a series of anchors various scales, aspect ratios angles, then the results can be outputted by performing once or twice classification bounding box regression for predefined anchors. However, most defined have relatively low accuracy, are useless following regression. In addition, preset...

With the successful application of convolutional neural network (CNN), significant progress has been made by CNN-based ship detection methods. However, they often face considerable difficulties when applied to a new domain where imaging condition changes significantly. Although training with two domains together can solve this problem some extent, large shift will lead sub-optimal feature representations, and thus weaken generalization ability on both domains. In paper, adaptive method is...

Infrared and visible images (multi-sensor or multi-band images) have many complementary features which can effectively boost the performance of object detection. Recently, convolutional neural networks (CNNs) seen frequent use to perform detection in images. However, it is very difficult for CNNs extract from infrared In order solve this problem, a difference maximum loss function proposed paper. The guide learning directions two base maximize between CNNs, so as diverse features. addition,...

Small ships in remote sensing images have blurred details and are difficult to detect. Existing algorithms usually detect small based on predefined anchors with different sizes. However, limited by the number of sizes, it is for anchor-based methods match sizes structures during training, as they can easily cause misdetections. In this paper, we propose a hybrid anchor structure generate region proposals ships, so take full advantage both high localization accuracy anchor-free fewer To unify...

Noninvertible symmetry generalizes traditional group symmetries, advancing our understanding of quantum matter, especially one-dimensional gapped systems. In critical lattice models, it is usually realized as emergent symmetries in the corresponding low-energy conformal field theories. this work, we study models with noninvertible Rep($D_8$) one dimension. This leads us to a new class points (QCP), symmetry-enriched QCPs, generalization known QCPs. They are phase transitions between...

Currently, reliable and accurate ship detection in optical remote sensing images is still challenging. Even the state-of-the-art convolutional neural network (CNN)-based methods cannot obtain very satisfactory results. To more accurately locate ships diverse orientations, some recent conduct via rotated bounding box. However, it further increases difficulty of because an additional variable orientation must be predicted algorithm. In this article, a novel CNN-based ship-detection method...

The impact of a drop on solid surface has been studied for many years. However, most the previous numerical simulations were focused at room temperature and standard atmospheric pressure. This paper presents study n-heptane n-decane drops impacting surfaces with consideration high pressure using smoothed particle hydrodynamics (SPH). SPH method is validated against experiments from our work literature. two typical drop-impact regimes, namely, spread rebound. Different sequences simulated...

In this paper, a novel single image super-resolution method unifying deep residual network and Wasserstein generative adversarial nets is proposed aiming at generating photo-realistic with finer texture details. Specifically, we construct framework consisting of generator that recovers high-resolution an input low-resolution discriminator tries to distinguish the recovered from real image. The competing drives produce images are highly similar images. Meanwhile, define new loss function by...

Cartesian Genetic Programming (CGP) is a powerful and popular tool for automatic generation of computer programs to solve user defined tasks. This paper proposes Co-evolutionary CGP (named Co-CGP) which can automatically gain high-order knowledge accelerate the search. In Co-CGP, two modules are working in cooperation given problem. One module focuses on solving series small scale problems same type generate building blocks. Simultaneously, second combing available blocks construct final...

In actual implementation, digital image retrieval are facing all kinds of problems. There still exists some difficulty in measures and methods for application. Currently there is not a unambiguous algorithm which can directly shown the obvious feature content satisfy color, scale invariance rotation simultaneously. So related technology about based on analyzed by us. The research focused global features such as seven HU invariant moments, edge direction histogram eccentricity. method blocked...

The abductive natural language inference task ($\alpha$NLI) is proposed to infer the most plausible explanation between cause and event. In $\alpha$NLI task, two observations are given, hypothesis asked pick out from candidates. Existing methods model relation each candidate separately penalize network uniformly. this paper, we argue that it unnecessary distinguish reasoning abilities among correct hypotheses; similarly, all wrong hypotheses contribute same when explaining reasons of...

Accurate ship segmentation in optical remote sensing images is challenging. In this paper, contour prediction network introduced to improve the performance of network. Ship predicts contour, which can promote learning network, producing more accurate results. The be naturally embedded into any with encoder-decoder architecture. Moreover, only used training process, and thus not accuracy, but also does increase computational cost testing process.