Recent studies show that the state-of-the-art deep neural networks (DNNs) are vulnerable to adversarial examples, resulting from small-magnitude perturbations added input. Given emerging physical systems using DNNs in safety-critical situations, examples could mislead these and cause dangerous situations. Therefore, understanding world is an important step towards developing resilient learning algorithms. We propose a general attack algorithm, Robust Physical Perturbations (RP2), generate...

There have been tremendous efforts and many technical innovations in supporting real-time video streaming the past two decades, but cost-effective large-scale broadcast has remained an elusive goal. Internet protocol (IP) multicast represented earlier attempt to tackle this problem failed largely due concerns regarding scalability, deployment, support for higher level functionality. Recently, peer-to-peer based emerged as a promising technique, which shown be cost effective easy deploy. This...

Deep neural networks (DNNs) are vulnerable to adversarial examples-maliciously crafted inputs that cause DNNs make incorrect predictions. Recent work has shown these attacks generalize the physical domain, create perturbations on objects fool image classifiers under a variety of real-world conditions. Such pose risk deep learning models used in safety-critical cyber-physical systems. In this work, we extend more challenging object detection models, broader class algorithms widely detect and...

Tracking maneuvering target in real time autonomously and accurately an uncertain environment is one of the challenging missions for unmanned aerial vehicles (UAVs). In this paper, aiming to address control problem tracking obstacle avoidance, online path planning approach UAV developed based on deep reinforcement learning. Through end-to-end learning powered by neural networks, proposed can achieve perception continuous motion output control. This includes: (1) A deterministic policy...

Knowledge graphs store a large number of factual triples while they are still incomplete, inevitably. The previous knowledge graph completion (KGC) models predict missing links between entities merely relying on fact-view data, ignoring the valuable commonsense knowledge. embedding (KGE) techniques suffer from invalid negative sampling and uncertainty link prediction, limiting KGC’s performance. To address above challenges, we propose novel scalable Commonsense-Aware Embedding (CAKE)...

Popular methods usually use a degradation model in supervised way to learn watermark removal model. However, it is true that reference images are difficult obtain the real world, as well collected by cameras suffer from noise. To overcome these drawbacks, we propose perceptive self-supervised learning network for noisy image (PSLNet) this paper. PSLNet depends on parallel remove noise and watermarks. The upper uses task decomposition ideas watermarks sequence. lower utilizes idea...

Recent advances in multi-radio multi-channel transmission technology have the potential of substantially improving system capacity multi-hop wireless networks. While previous work has primarily focused on link level protocol design, this paper we investigate achievable performance gain, by jointly optimizing routing and scheduling a multichannel network. We formulate optimization under deterministic model, seek to minimize overall activation time use satisfy given end-to-end traffic demands...

One of the distinctive features in a wireless ad hoc network is lack any central controller or single point authority, which each node/link then makes its own decisions independently. Therefore, fully cooperative behaviors, such as cooperation for increasing system capacity, mitigating interference other, honestly revealing private information, might not be directly applied. It has been shown that power control an efficient approach to achieve quality service (QoS) requirement networks....

In recent years, there has been significant interest in adopting the peer-to-peer (P2P) technology for Internet live video streaming. There are primarily two reasons behind this development: elimination of infrastructure support and self-scaling property P2P systems. The success our system Coolstreaming represented one earliest large-scale streaming experiments. Since then, have several commercial deployments. With desirable content, these systems potential to scale orders magnitude beyond...

Voice activity detection (VAD) is the task of predicting which parts an utterance contains speech versus background noise. It important first step to determine samples send decoder and when close microphone. The long short-term memory neural network (LSTM) a popular architecture for sequential modeling acoustic signals, has been successfully used in several VAD applications. However, it observed that LSTMs suffer from state saturation problems (i.e., voice dictation tasks), thus requires...

This paper combines deep reinforcement learning (DRL) with meta-learning and proposes a novel approach, named meta twin delayed deterministic policy gradient (Meta-TD3), to realize the control of unmanned aerial vehicle (UAV), allowing UAV quickly track target in an environment where motion is uncertain. approach can be applied variety scenarios, such as wildlife protection, emergency aid, remote sensing. We consider multi-task experience replay buffer provide data for DRL algorithm, we...

A survey on automated lip-reading approaches is presented in this paper with the main focus being deep learning related methodologies which have proven to be more fruitful for both feature extraction and classification. This also provides comparisons of all different components that make up systems including audio-visual databases, extraction, classification networks schemas. The contributions unique insights are: 1) comparison Convolutional Neural Networks other neural network architectures...

In this paper, an intelligent algorithm integrating model predictive control and Standoff is proposed to solve trajectory planning that UAVs may face while tracking a moving target cooperatively in complex three-dimensional environment. A fusion using thus constructed ensure formation maintenance, maximizing UAV sensors’ detection range minimizing loss probability. Meanwhile, with model, fully connected communication topology used complete the communication, multi-UAV can be reconfigured...

Abstract This paper is concerned with parameter estimation of Wiener systems measurement noises employing correlation analysis method and adaptive Kalman filter. The presented system consists two series blocks, that is, a dynamic block represented by auto‐regressive moving average (ARMA) model, static nonlinear established neural fuzzy model. Aim at estimating separately the separable signals are introduced. First, applying to decouple identification linear from block, then ARMA model...

Recently, IEEE drew up a new task group named TGax to draft out the standard 802.11ax for next generation WLANs. However, average throughput is very low due current medium access control (MAC) protocol. A promising solution this problem draw Orthogonal Frequency Division Multiple Access (OFDMA) into enable multiuser access. The key challenges of adopting OFDMA are synchronization and overhead reduction. In paper, we propose an based (OMAX) protocol solve both two above. whole channel...