首页> 外国专利> Being neural network study

Being neural network study

机译:正在进行神经网络研究

摘要

PROBLEM TO BE SOLVED: To provide an air conditioner for performing air-conditioning control by a neural network capable of reducing a learning error and contracting a learning time simultaneously in initializing the neural network. SOLUTION: A region, where a user is most likely to recognize a change of air-conditioning control quantity such as a conditioned air temperature, airflow quantity, and supply opening mode, is set, and a weighting factor W of teacher data in the region is set to be large. Thereby, the learning error in the region where the user is most likely to recognize the change of the air- conditioning control quantity is preferentially decreased as compared with the leaning error in other region so that reduction of the learning error and contraction of the learning time can be simultaneously achieved.
机译:要解决的问题:提供一种用于通过神经网络执行空调控制的空调,该空调能够在初始化神经网络的同时减少学习错误并缩短学习时间。解决方案:设置了一个区域,在该区域中,用户最有可能识别空调控制量的变化,例如空调温度,气流量和供气打开方式,并且该区域中教师数据的加权因子W设置为大。从而,与其他区域中的倾斜误差相比,优选地减少了用户最有可能认识到空调控制量的变化的区域中的学习误差,从而减小了学习误差并缩短了学习时间。可以同时实现。

著录项

  • 公开/公告号JP3826763B2

    专利类型

  • 公开/公告日2006-09-27

    原文格式PDF

  • 申请/专利权人 株式会社デンソー;

    申请/专利号JP20010331255

  • 发明设计人 一志 好則;立石 雅彦;

    申请日2001-10-29

  • 分类号B60H1;B60H1/32;G05B11/32;G05B13/02;G06N3;G06N3/08;

  • 国家 JP

  • 入库时间 2022-08-21 21:50:46

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号