ID 原文 译文
56478 该架构包括通信环境理解、通信波形适配和智能节点学习进化3个核心功能,以及支持这些功能的通信计算融合硬件平台. The architecture includes three core functions: communi?cation environment understanding, communication waveform adaptation, and learning and evolution.
56479 所提出的智能适变架构支持通信环境知识库、通信波形库,以及波形与环境适配知识图谱的不断累积和进化,通过波形在线重构,通信节点既能匹配典型通信场景,又能快速适应未知环境,因而支持智能通信节点的可持续发展. To support these functions, it also provides a hardware platform that integrates the capabilities of communication and computation. The proposed intelligent adaptive architecture supports the continuous accumulation and evolution of knowledge bases of communication environments, communication waveforms, and the match between them.
56480 进一步本文梳理了强化学习、在线学习和迁移学习等3种机器学习技术在智能适变无线通信节点中的应用,并以最经典的信道估计过程为代表,给出了机器学习应用于通信环境识别的典型范例. Through online waveform reconfiguration, communication nodes can adapt to typical communication scenarios and unknown environments, thus support the sustainable development of intelligent communication. Further?more, this paper summarizes the applications of reinforcement learning, online learning, and transfer learning in intelligent adaptive wireless communications, also providing a typical example of the application of machine learning to the channel estimation process.
56481 在很多真实应用中,数据以流的形式不断被收集得到. In many real-world applications, data are collected in the form of streams.
56482 由于数据收集环境往往发生动态变化,流数据的分布也会随时间不断变化. As a result of the evolvingnature of dynamic environments, the distribution of data streams generally changes over time.
56483 传统的机器学习技术依赖于数据独立同分布假设,因而在这类分布变化的流数据学习问题上难以奏效. Such distributionchanges hinder the application of conventional machine learning approaches because the fundamental assumptionof independent and identical distribution does not hold in these scenarios.
56484 本文提出一种基于决策树模型重用的算法进行分布变化的流数据学习. This paper proposes an algorithmbased on the decision tree model reuse mechanism for learning from distribution-changing data streams.
56485 该算法是一种在线集成学习方法:算法将维护一个模型库,并通过决策树模型重用机制更新模型库. Theproposed algorithm is essentially an online ensemble method that maintains a model pool and updates it byperforming decision tree model reuse.
56486 其核心思想是希望从历史数据中挖掘与当前学习相关的知识,从而抵御分布变化造成的影响. The main idea is to exploit the useful knowledge in historical data to helpresist the negative effects of distribution changes.
56487 通过在合成数据集和真实数据集上进行实验,我们验证了本文提出方法的有效性. We validate the effectiveness of the proposed approach throughexperiments on synthetic and real-world datasets.