Despite the impressive performance in many individual tasks, deep neural networks suffer from catastrophic forgetting when learning new tasks incrementally. Recently, various incremental methods have been proposed, and some approaches achieved acceptable relying on stored data or complex generative models. However, storing previous is limited by memory privacy issues, models are usually unstable inefficient training. In this paper, we propose a simple non-exemplar based method named PASS, to...

In recent years, pre-trained models have become dominant in most natural language processing (NLP) tasks. However, the area of Automated Essay Scoring (AES), such as BERT not been properly used to outperform other deep learning LSTM. this paper, we introduce a novel multi-scale essay representation for that can be jointly learned. We also employ multiple losses and transfer from out-of-domain essays further improve performance. Experiment results show our approach derives much benefit joint...

An essentialtask of the electronic support measures system is to deinterleave intercepted radar pulse stream. To improve radars' antijamming capability and adapt radars different application scenarios, signals emitted by various are becoming abundant, causing stream reconnaissancereceiver becomemore more complicated. Apart from fixed repetition interval (PRI) signals, other PRI modulated such as jitter staggered PRI, also appear in The aforementioned defined mixed this article....

Deterministic performance is a key enabler for 5G networking. In this context, we present highly scalable Large-scale Network (LDN) architecture providing end-to-end latency and bounded jitter guarantees in IP networks. At the data plane, flows are first shaped at ingress gateways using gate-control queues, achieving very fine granularity compared to existing state of art solutions. Inside network, traffic scheduled an asynchronous cyclic queuing mechanism that can be implemented real...

The Kaczmarz method, or the algebraic reconstruction technique (ART), is a popular method for solving large-scale overdetermined systems of equations. Recently, Strohmer et al. proposed randomized algorithm, an improvement that guarantees exponential convergence to solution. This has spurred much interest in algorithm and its extensions. We provide this paper exact formula mean squared error (MSE) value reconstructed by algorithm. also compute decay rate MSE, which we call "annealed"...

In this work, we fabricate uniform wrinkles on ITO and systematically study the properties of wrinkled in optics, electrics mechanics. The shows a high optical transmittance improved mechanical bending performance.

Abstract Complex surface structures have stimulated a great deal of interests due to many potential applications in devices. However, the fabrication complex micro-/nanostructures, there are always challenges precise design, or good controllability, low cost, high throughput. Here, we present route for accurate design and highly controllable quasi-three-dimensional (quasi-3D) based on thermal deformation simple two-dimensional laser-induced patterns. A quasi-3D structure, coaxially nested...

Characterization of corrosion products formed on tin-bronze exposed to industry atmosphere for a 29-year period was investigated. Results demonstrated that the rate 0.56 μm·a−1. Tin-bronze had two completely different structures exposure sides—the layer skyward side uneven with cracks, and product field-ward is double-layer structure dense inner porous outer layer. Sn uniformly distributed in side, densely concentrated near substrate side.

Tensor renormalization group (TRG) method is a real space approach. It has been successfully applied to both classical and quantum systems. In this paper, we study disordered frustrated system, the two-dimensional Edwards-Anderson model, by new topological invariant TRG scheme. We propose an approach calculate local magnetizations nearest pair correlations simultaneously. The Nishimori multicritical point predicted agrees well with recent Monte Carlo results. schemes outperform mean-field...

An efficient algorithm based on the frequency scaling (FSA) for terahertz (THz) stepped-frequency (SF) synthetic aperture radar (SAR) imaging is proposed in this article. THz SF SAR has an urgent application demand close-range medical imaging, security check, and near-space military target reconnaissance. In these fields, real-time performance of great significance. The back projection (BPA) accurate time-domain algorithm, but it a huge amount calculation cannot meet requirements imaging....

An efficient algorithm based on chirp scaling (CSA) for terahertz (THz) stepped-frequency (SF) synthetic aperture radar (SAR) imaging is proposed in this letter. We call SF-CSA convenience. THz techniques have good prospects the fields of military, public safety, and nondestructive defect identification. In these fields, real-time very important. CSA an suitable SAR imaging, but it can be only used linear frequency-modulation (LFM) imaging. The main contribution letter successful application...

Recently, orthogonal matching pursuit (OMP) has been widely used in direction of arrival (DOA) studies, which not only greatly improves the resolution DOA, but can also be applied to single-snapshot and coherent source cases. When applying OMP algorithm rectangular array DOA millimeter-wave radar, it is necessary reshape two-dimensional (2D) signal into a long one-dimensional (1D) signal. However, 1D will increase number length atoms complete dictionary algorithm, amount computation. Taking...

The current system to manage the global pool of IP addresses is centralized in five transnational organizations, Regional Internet Registries (RIRs). Each these RIRs address for a large number countries. Because are private they subject legal framework country where based. This configuration results jurisdictional overflow from countries RIR based all that serving (the served by de facto become subjects hosted). situation aggravated deployment new security techniques such as RPKI and BGPsec,...

Software-Defined Networking (SDN) is an emerging architecture that ideal for today's high-bandwidth, dynamic network environments. In this architecture, the control and data planes are decoupled from each other. Although much research has been performed into how SDN can resolve some of most-glaring security issues traditional networking, less addressed cloud threats, and, in particular, botnet/malware detection in-cloud attacks. This work proposes intrusion prevention system networking with...

We demonstrate through two case studies, one on the p-spin interaction model and other random K-satisfiability problem, that a heterogeneity transition occurs to ground-state configuration space of finite-connectivity spin glass system at certain critical value constraint density. At point, exponentially many communities emerge from space, making entropy density s(q) configuration-pairs non-concave function configuration-pair overlap q. Each community is collection relatively similar...

The cluster variation method has been developed into a general theoretical framework for treating short-range correlations in many-body systems after it was first proposed by Kikuchi 1951. On the numerical side, message-passing approach called generalized belief propagation (GBP) Yedidia, Freeman and Weiss about decade ago as way of computing minimal value variational free energy marginal distributions clusters variables. However GBP equations are often redundant, is quite non-trivial task...