We present a novel multi-agent RL approach, Selective Multi-Agent Prioritized Experience Relay, in which agents share with other agents a limited number of transitions they observe during training. The intuition behin...
We present a novel multi-agent RL approach, Selective Multi-Agent Prioritized Experience Relay, in which agents share with other agents a limited number of transitions they observe during training. The intuition behind this is that even a small number of relevant experiences from other agents could help each agent learn. Unlike many other multi-agent RL algorithms, this approach allows for largely decentralized training, requiring only a limited communication channel between agents. We show that our approach outperforms baseline no-sharing decentralized training and state-of-the art multi-agent RL algorithms. Further, sharing only a small number of highly relevant experiences outperforms sharing all experiences between agents, and the performance uplift from selective experience sharing is robust across a range of hyperparameters and DQN variants.
Detecting suspicious activities in public places with higher people gathering and interaction has turned out to be an act with growing interest due to the increasing number of crime scenes and causalities happening in...
Detecting suspicious activities in public places with higher people gathering and interaction has turned out to be an act with growing interest due to the increasing number of crime scenes and causalities happening in these days. Surveying and tracking of human activities are increasingly difficult owing to the random nature of human movements and actions. The reliability is greatly affected due to this randomness. Also a human operator cannot continuously monitor multiple screens efficiently in a consequent manner so an automated surveillance system deployment becomes a necessity. Currently, tracking individuals may be done remotely, and the analysis of the recorded images can be automated using object detection models, with the help of high resolution cameras and the development of machine learning techniques. This proposed system aims in identifying threats that are probable to occur in a public gathering or space which may be an explosion, accident or possession of armoury, etc. This proposed model takes advantage of the information from the image data to learn complex patterns and develop pattern recognition technique to identify the anomalies using high resolution camera and alert the monitoring authority in order to take the necessary actions. This proposed work compares various object detection techniques of machine learning algorithms and suggests the best model based on its performance metrics.
Bragg-type layered dielectric structure, which provides the Chebyshev character of the behavior of the reflection frequency response, was considered. To obtain a fractional-rational representation of the structure fre...
详细信息
Magnetic nanoparticles can be embedded in electrospun nanofibers and other polymeric matrices to prepare magnetic composites with defined magnetic and mechanical properties. Metal-oxide nanoparticles, such as magnetit...
详细信息
In the early 21st century, Internet of Drones (IoD) ushered in a rapid growth period and is gradually expanded from military field to civilian field. Currently, there are more and more scenarios where multiple drones ...
详细信息
Big data and big data analytics have been used in various types of businesses and organizations. Higher education institutions (HEIs) produce and process large amounts of different types of data that satisfy big data ...
详细信息
ISBN:
(数字)9798350352962
ISBN:
(纸本)9798350352979
Big data and big data analytics have been used in various types of businesses and organizations. Higher education institutions (HEIs) produce and process large amounts of different types of data that satisfy big data properties. However, HEIs have not yet reaped the highest benefits out of big data due to several challenges including the lack of a comprehensive framework for using big data in higher education. In this paper, we pave the way for proposing such a framework starting with a detailed review and comparisons of the latest big data in higher education frameworks. Frameworks are compared based on specific criteria to identify the weaknesses and discover the gaps that we can overcome in the proposed framework.
Amazing results in many works in the field of deep convolutional neural networks using them in many fields of machine learning such as image classification control of atrial play, and image annotation. Mathematical an...
详细信息
Pseudo-Boolean optimization (PBO) is usually used to model combinatorial optimization problems, especially for some real-world applications. Despite its significant importance in both theory and applications, there ar...
详细信息
In this paper, we explore the potential of deep learning techniques in the field of ultra-fast laser processing. More specifically, we trained convolutional neural networks on an in-house dataset with the aim of predi...
详细信息
We present new algorithms for the k mismatches version of approximatestring matching. Our algorithms utilize the SIMD (Single Instruction MultipleData) instruction set extensions, particularly AVX2 and AVX-512 *** app...
详细信息
暂无评论