股票市场具有变化快、干扰因素多、周期数据不足等特点,股票交易是一种不完全信息下的博弈过程,单目标的监督学习模型很难处理这类序列化决策问题。强化学习是解决该类问题的有效途径之一。提出了基于深度强化学习的智能股市操盘手模型I...股票市场具有变化快、干扰因素多、周期数据不足等特点,股票交易是一种不完全信息下的博弈过程,单目标的监督学习模型很难处理这类序列化决策问题。强化学习是解决该类问题的有效途径之一。提出了基于深度强化学习的智能股市操盘手模型ISTG(Intelligent Stock Trader and Gym),融合历史行情数据、技术指标、宏观经济指标等多数据类型,分析评判标准和优秀控制策略,加工长周期数据,实现可增量扩展不同类型数据的复盘模型,自动计算回报标签,训练智能操盘手,并提出直接利用行情数据计算单步确定性动作值的方法。采用中国股市1400多支的有10年以上数据的股票进行多种对比实验,ISTG的总体收益达到13%,优于买入持有总体−7%的表现。展开更多
In this paper, the performance of frequency synchronization in a multiple-input multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) system is analyzed for the purpose of carrier frequency offset ...In this paper, the performance of frequency synchronization in a multiple-input multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) system is analyzed for the purpose of carrier frequency offset (CFO) estimation and compensation. Specifically, a joint transmit antenna selection (ST) and receive maximum ratio combining (MRC) (ST/MRC) method is adopted, that is, only one transmit antenna with the highest channel power is selected while MRC is used at the receiver to maximize the sum of frequency synchronization metric of CFO estimation are derived for several antenna configurations validate the theoretical analysis. The mean square error (MSE) closed-form expressions Simulations in both flat and multipath fading channels展开更多
文摘股票市场具有变化快、干扰因素多、周期数据不足等特点,股票交易是一种不完全信息下的博弈过程,单目标的监督学习模型很难处理这类序列化决策问题。强化学习是解决该类问题的有效途径之一。提出了基于深度强化学习的智能股市操盘手模型ISTG(Intelligent Stock Trader and Gym),融合历史行情数据、技术指标、宏观经济指标等多数据类型,分析评判标准和优秀控制策略,加工长周期数据,实现可增量扩展不同类型数据的复盘模型,自动计算回报标签,训练智能操盘手,并提出直接利用行情数据计算单步确定性动作值的方法。采用中国股市1400多支的有10年以上数据的股票进行多种对比实验,ISTG的总体收益达到13%,优于买入持有总体−7%的表现。
基金supported by the National Science and Technology Major Project of the Ministry of Science and Technology (2015ZX03002008)the National Natural Science Foundation of China (61322110)
文摘In this paper, the performance of frequency synchronization in a multiple-input multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) system is analyzed for the purpose of carrier frequency offset (CFO) estimation and compensation. Specifically, a joint transmit antenna selection (ST) and receive maximum ratio combining (MRC) (ST/MRC) method is adopted, that is, only one transmit antenna with the highest channel power is selected while MRC is used at the receiver to maximize the sum of frequency synchronization metric of CFO estimation are derived for several antenna configurations validate the theoretical analysis. The mean square error (MSE) closed-form expressions Simulations in both flat and multipath fading channels