摘要
随着计算机硬件能力与软件算法的发展,带有大量隐藏层的深度神经网络技术取得了突破性的进展,它对非线性问题的处理能力,使得其在各行各业都有极为广阔的应用前景。它作为计算机科学的一个前沿领域,爆炸式地发展出了大量具有不同应用指向的分支,对于广大建筑师而言显然是陌生的。文章立足国际主流深度神经网络的研究成果,以建筑师的应用视角,即以建筑实践与研究中经常遇到的数据类型与问题类型出发,介绍了其输入与输出的数据描述,以及基本的网络模型架构选择原理。
With the development of computer hardware and algorithms, deep neural network technology with a large number of hidden layers has made a significant breakthrough in dealing with nonlinear problems that is applicable in all aspects of our daily lives. As a frontier field of computer science,it has explosively developed into a large number of sub-models in various applications, which is obviously interesting and unfamiliar to architects. This study tries to introduce its variable interfaces and basic principles in sub model selection, from an architectural perspective, concerning daily data types and problem typologies in architectural practices and researches.
出处
《时代建筑》
2018年第1期50-55,共6页
Time + Architecture
基金
国家重点研发计划项目<南方地区城镇居住建筑绿色设计新方法与技术协同优化>(2016YFC0700207)
国家自然科学基金面上项目<公共建筑中人流模拟的"互联网+"方法>(51778417)
关键词
人工智能
机器学习
设计数据
设计问题
模型遴选
Artificial Intelligence
Machine Learning
Design Data
Design Problem
Model Selection
