Click on the top " AI proper way ", select the "star" Public No.
Heavy dry goods, the first time served
Neural Networks
Since the breakthrough in 2012 CNN's imagenet, neural network-based network began to rage deep learning academics and industry. Let's look at a picture, the number of google internal depth study on the project. And a very wide field of application, from Android to find drugs to youtube.
The amount of our past lives under review from the neural network together:
• 1958: Perceptron (linear model)
• 1969: Perceptron has limitation
• 1980s: Multi-layer perceptron
• Do not have significant difference from DNN today
• 1986: Backpropagation
• Usually more than 3 hidden layers is not helpful
• 1989: 1 hidden layer is “good enough”, why deep?
• 2006: RBM initialization (breakthrough)
• 2009: GPU
• 2011: Start to be popular in speech recognition
• 2012: win ILSVRC image competition
Deep learning is a branch of machine learning, to say what is currently the most important branch. Learn how to learn the depth of some of it?
But it is still crucial three steps:
1. Select the neural network
2. Define the quality of neural networks
3. Select the best set of parameters
The following is a diagram of neural networks:
And θ b are all within neurons
1. fully connected network (Fully Connection)
2. The depth of the network DEEP
Many layer depth =
Then someone will ask:
* 到底多少层深度合适?每层多个神经元?
答:这个看经验和实验的结果,不断调整。
* 结构能被自动设定吗?
答:可以通过进化网络实现。
* 我们能自己设计网络结构吗?
答: CNN 就是设计出来的网络结构。
3. 定义神经网络的好坏Loss
我们以minist 数字识别为例,一组数字识别为例
4. 选择最好的神经网络(找到参数集)
核心方法:
* Gradient Descent
* BackPropagation
深度学习基本知识点了解到了,但是为什么越Deep,效果会越好? 以前都是做类比思考,比如电路模型,但是近期的lpaper上在理论上有严格的证明,我们后续博客会介绍
本专栏图片、公式很多来自台湾大学李弘毅老师、斯坦福大学cs229、cs231n 、斯坦福大学cs224n课程。在这里,感谢这些经典课程,向他们致敬!
作者简介:武强 兰州大学博士,谷歌全球开发专家Google Develop Expert(GDE Machine Learing 方向)
CSDN:https://me.csdn.net/dukuku5038
知乎:https://www.zhihu.com/people/Dr.Wu/activities
漫画人工智能公众号:DayuAI-Founder
推荐阅读
(点击标题可跳转阅读)
重磅!
林轩田机器学习完整视频和博主笔记来啦!
扫描下方二维码,添加 AI有道小助手微信,可申请入群,并获得林轩田机器学习完整视频 + 博主红色石头的精炼笔记(一定要备注:入群 + 地点 + 学校/公司。例如:入群+上海+复旦。
长按扫码,申请入群
(添加人数较多,请耐心等待)
最新 AI 干货,我在看