随着智能家居应用的不断深化,基于Wi-Fi信号的室内定位技术也受到了广泛关注。在实际应用中,大多数室内定位算法采集得到的训练数据和测试数据通常并非来自于同一理想环境,各种环境条件变化以及信号漂移导致采集得到的训练数据和测试数...随着智能家居应用的不断深化,基于Wi-Fi信号的室内定位技术也受到了广泛关注。在实际应用中,大多数室内定位算法采集得到的训练数据和测试数据通常并非来自于同一理想环境,各种环境条件变化以及信号漂移导致采集得到的训练数据和测试数据间的概率分布不同。传统定位模型在面对不同分布的训练数据和测试数据时无法保证具有良好的定位精度,常出现算法定位精度大幅降低,甚至算法不可用等问题。面对这一难点,迁移学习中的域适应方法作为一种可以有效解决训练样本和测试样本概率分布不一致的学习问题被广泛应用于室内定位领域。文中结合域适应学习和机器学习算法,提出了一种基于特征迁移的室内定位算法(Transfer Learning Location AlgorithmBased on Global and Local Metrics Adaptation,TL-GLMA)。TL-GLMA在定位阶段通过特征迁移方式将两域原始数据映射至高维空间,从而在最小化两域数据的分布差异的同时保留两域数据内部的局部几何属性,并利用映射后的独立同分布数据训练分类器,从而实现目标定位。实验结果表明,TL-GLMA能够有效减少环境变化带来的干扰,提升定位精度。展开更多
TFP growth may derive from both technology progress(technical effect) and factor allocation(structural effect). Using China's macroeconomic and industrial data, this paper decomposes China's TFP growth on the ...TFP growth may derive from both technology progress(technical effect) and factor allocation(structural effect). Using China's macroeconomic and industrial data, this paper decomposes China's TFP growth on the basis of growth accounting to cast light on China's growth sources since reform and opening up in 1978. Our study has led to the following findings:(1) From 1978 to 2014, China's economic growth was of generally good quality, and about 1/3 of growth momentum stemmed from a general technology improvement.(2) After 2005, China's late-mover advantage diminished due to narrowed technology gaps with advanced economies. This resulted in a sharp decline in the contribution of technology progress to growth. However, structural effect contributed a steadily increasing share to China's growth.(3) After global financial crisis in 2008, there has been a tendency of reverse technology progress in terms of factor allocation in sectors with excess industrial capacity and other sectors like finance and real estate. Therefore, China should divert its factor resources to more tech-intensive and efficient sectors in the short run, and strive to promote technology progress in all sectors in a longer timeframe.展开更多
文摘随着智能家居应用的不断深化,基于Wi-Fi信号的室内定位技术也受到了广泛关注。在实际应用中,大多数室内定位算法采集得到的训练数据和测试数据通常并非来自于同一理想环境,各种环境条件变化以及信号漂移导致采集得到的训练数据和测试数据间的概率分布不同。传统定位模型在面对不同分布的训练数据和测试数据时无法保证具有良好的定位精度,常出现算法定位精度大幅降低,甚至算法不可用等问题。面对这一难点,迁移学习中的域适应方法作为一种可以有效解决训练样本和测试样本概率分布不一致的学习问题被广泛应用于室内定位领域。文中结合域适应学习和机器学习算法,提出了一种基于特征迁移的室内定位算法(Transfer Learning Location AlgorithmBased on Global and Local Metrics Adaptation,TL-GLMA)。TL-GLMA在定位阶段通过特征迁移方式将两域原始数据映射至高维空间,从而在最小化两域数据的分布差异的同时保留两域数据内部的局部几何属性,并利用映射后的独立同分布数据训练分类器,从而实现目标定位。实验结果表明,TL-GLMA能够有效减少环境变化带来的干扰,提升定位精度。
基金supported by CASS Innovation Project“Analysis and Estimation of Innovation-Driven Development”(10620161001005)National Soft Science Program“Industrial Structural Transition,Technology Innovation and Improvement of China’s Economic Growth Potentials”(2014GXS4B073)Center of the Research of Chinese Socialism Theories/Major Program of the National Social Science Foundation of China(NSFC)“Study on Innovation-Driven Development Strategy and Mass Entrepreneurship and Mass Innovation”(2015YZD03)
文摘TFP growth may derive from both technology progress(technical effect) and factor allocation(structural effect). Using China's macroeconomic and industrial data, this paper decomposes China's TFP growth on the basis of growth accounting to cast light on China's growth sources since reform and opening up in 1978. Our study has led to the following findings:(1) From 1978 to 2014, China's economic growth was of generally good quality, and about 1/3 of growth momentum stemmed from a general technology improvement.(2) After 2005, China's late-mover advantage diminished due to narrowed technology gaps with advanced economies. This resulted in a sharp decline in the contribution of technology progress to growth. However, structural effect contributed a steadily increasing share to China's growth.(3) After global financial crisis in 2008, there has been a tendency of reverse technology progress in terms of factor allocation in sectors with excess industrial capacity and other sectors like finance and real estate. Therefore, China should divert its factor resources to more tech-intensive and efficient sectors in the short run, and strive to promote technology progress in all sectors in a longer timeframe.
