Communication could potentially be an effective way for multi-agent cooperation. However, information sharing among all agents or in predefined communication architectures that existing methods adopt can problematic. When there is a large number of agents, cannot differentiate valuable helps cooperative decision making from globally shared information. Therefore, barely helps, and even impair the learning Predefined architectures, on other hand, restrict thus restrain potential To tackle...

Background: Despite recent significant progress in the development of automatic sleep staging methods, building a good model still remains big challenge for studies with small cohort due to data-variability and data-inefficiency issues. This work presents deep transfer learning approach overcome these issues enable transferring knowledge from large dataset staging. Methods: We start generic end-to-end framework sequence-to-sequence derive two networks as means learning. The are first trained...

Learning to cooperate is crucially important in multi-agent environments. The key understand the mutual interplay between agents. However, environments are highly dynamic, where agents keep moving and their neighbors change quickly. This makes it hard learn abstract representations of To tackle these difficulties, we propose graph convolutional reinforcement learning, convolution adapts dynamics underlying environment, relation kernels capture by representations. Latent features produced...

Community detection is an important issue due to its wide use in designing network protocols such as data forwarding Delay Tolerant Networks (DTN) and worm containment Online Social (OSN). However, most of the existing community algorithms focus on binary networks. Since networks are naturally weighted DTN or OSN, this article, we address problems networks, exploit for demonstrate how can facilitate these designs. Specifically, propose a novel algorithm, introduce two metrics:...

Lightweight models are pivotal in efficient semantic segmentation, but they often suffer from insufficient context information due to limited convolution and small receptive field. To address this problem, we propose a tailored approach segmentation by leveraging two complementary distillation schemes for supplementing networks: 1) self-attention scheme, which transfers long-range knowledge adaptively large teacher networks student networks; 2) layer-wise structured deep layers shallow...

This paper proposes SymDetector, a smartphone based application to unobtrusively detect the sound-related respiratory symptoms occurred in user's daily life, including sneeze, cough, sniffle and throat clearing. SymDetector uses built-in microphone on continuously monitor acoustic data multi-level processes classify symptoms. Several practical issues are considered developing such as users' privacy concerns about their data, resource constraints of different contexts smartphone. We have...

In this paper, we investigate how to network smart-phones for providing communications in disaster recovery. By bridging the gaps among different kinds of wireless networks, have designed and implemented a system called TeamPhone, which provides smartphones capabilities Specifically, TeamPhone consists two components: messaging self-rescue system. The integrates cellular networking, ad-hoc networking opportunistic seamlessly, enables rescue workers. energy-efficiently groups trapped survivor...

Traffic signal control aims to coordinate traffic signals across intersections improve the efficiency of a district or city. Deep reinforcement learning (RL) has been applied recently and demonstrated promising performance where each is regarded as an agent. However, there are still several challenges that may limit its large-scale application in real world. On one hand, policy current often heavily influenced by neighbor agents, coordination between agent neighbors needs be considered....

In this paper, we investigate how to network smartphones for providing communications in disaster recovery. By bridging the gaps among different kinds of wireless networks, have designed and implemented a system called TeamPhone, which provides capabilities Specifically, TeamPhone consists two components: A messaging self-rescue system. The integrates cellular networking, ad-hoc opportunistic networking seamlessly, enables rescue workers. groups, schedules, positions trapped survivors. Such...

The emerging of mobile social networks opens opportunities for viral marketing. However, before fully utilizing as a platform marketing, many challenges have to be addressed. In this paper, we address the problem identifying small number individuals through whom information can diffused network soon possible, referred diffusion minimization problem. Diffusion under probabilistic model formulated an asymmetric k-center which is NP-hard, and best known approximation algorithm has ratio log...

Opportunistic mobile networks consisting of intermittently connected devices have been exploited for various applications, such as computational offloading and mitigating cellular traffic load. Different from existing work, in this paper, we focus on cooperatively data among to maximally improve the probability delivery a device an remote server or center within given time constraint, which is referred cooperative problem. Unfortunately, NP-hard. To end, heuristic algorithm designed based...

Communication lays the foundation for human cooperation. It is also crucial multi-agent However, existing work focuses on broadcast communication, which not only impractical but leads to information redundancy that could even impair learning process. To tackle these difficulties, we propose Individually Inferred (I2C), a simple yet effective model enable agents learn prior agent-agent communication. The knowledge learned via causal inference and realized by feed-forward neural network maps...

Offline reinforcement learning (RL) defines the task of from a static logged dataset without continually interacting with environment. The distribution shift between learned policy and behavior makes it necessary for value function to stay conservative such that out-of-distribution (OOD) actions will not be severely overestimated. However, existing approaches, penalizing unseen or regularizing policy, are too pessimistic, which suppresses generalization hinders performance improvement. This...

Achieving human-level dexterity is an important open problem in robotics. However, tasks of dexterous hand manipulation, even at the baby level, are challenging to solve through reinforcement learning (RL). The difficulty lies high degrees freedom and required cooperation among heterogeneous agents (e.g., joints fingers). In this study, we propose Bimanual Dexterous Hands Benchmark (Bi-DexHands), a simulator that involves two hands with tens bimanual manipulation thousands target objects....

Dynamic textures are sequences of images moving scenes that exhibit certain stationarity properties in time, for example, sea-waves, smoke, foliage, whirlwind etc. This work proposed a novel characterization dynamic poses the problems recognizing. A method by spatio-temporal multiresolution histogram based on velocity and acceleration fields is presented. The has many desirable including simple computing, spatial efficiency, robustness to noise ability encoding information, which can...

Community detection is an important issue due to its wide use in designing network protocols such as data forwarding Delay Tolerant Networks (DTN) and worm containment Online Social (OSN). However, most of the existing community algorithms focus on binary networks. Since networks are weighted social networks, DTN or OSN, this paper, we address problems exploit for OSN. We propose a novel algorithm, then introduce two metrics called intra-centrality inter-centrality, characterize nodes...

The emerging of mobile social networks opens opportunities for viral marketing. However, before fully utilizing as a platform marketing, many challenges have to be addressed. In this paper, we address the problem identifying small number individuals through whom information can diffused network soon possible, referred <i>diffusion minimization</i> problem. Diffusion minimization under probabilistic diffusion model formulated an asymmetric <inline-formula> <tex-math...

Deep reinforcement learning (RL) has been applied to traffic signal control recently and demonstrated superior performance conventional methods. However, there are still several challenges we have address before fully applying deep RL control. Firstly, the objective of is optimize average travel time, which a delayed reward in long time horizon context RL. existing work simplifies optimization by using queue length, waiting delay, etc., as immediate presumes these short-term targets always...

Transformer has been considered the dominating neural architecture in NLP and CV, mostly under supervised settings. Recently, a similar surge of using Transformers appeared domain reinforcement learning (RL), but it is faced with unique design choices challenges brought by nature RL. However, evolution RL not yet well unraveled. In this paper, we seek to systematically review motivations progress on RL, provide taxonomy existing works, discuss each sub-field, summarize future prospects.

Convolutional Neural Networks (CNNs) have revolutionized the research in computer vision, due to their ability capture complex patterns, resulting high inference accuracies. However, increasingly nature of these neural networks means that they are particularly suited for server computers with powerful GPUs. We envision deep learning applications will be eventually and widely deployed on mobile devices, e.g., smartphones, self-driving cars, drones. Therefore, this paper, we aim understand...

Introduction: Acute lung injury (ALI) and its most severe form acute respiratory distress syndrome (ARDS) are commonly occurring devastating conditions that seriously threaten the system in critically ill patients. The current treatments improve oxygenation patients with ALI/ARDS short term, but do not relieve clinical mortality of ARDS. Purpose: To develop novel drug delivery systems can enhance therapeutic efficacy impede adverse effects drugs. Methods: Based on key pathophysiological...

Recently, there are many efforts attempting to learn useful policies for continuous control in visual reinforcement learning (RL). In this scenario, it is important a generalizable policy, as the testing environment may differ from training environment, e.g., exist distractors during deployment. Many practical algorithms proposed handle problem. However, best of our knowledge, none them provide theoretical understanding what affects generalization gap and why their methods work. paper, we...