Convolutional Neural Network
- Use Python to impliment a simple network for hanndwritten numeral classification.
- At some point in your daily life,you may have seen some practical application of the target recognition algorithm,such as face detection on your mobile phone camera.But what is the principle of its work?The core of these computer vision solutions is the convolution neural network(CNN).Simply speaking,these networks are particularly good at constructing complex features based on less complex features.A classic example is the face detector.The face detector.The early layers are responsible for identifying vertical and horizontal lines,and the later stage is responsible for finding the nose and mouth.
Primary neural network
This article wil not introduce the general working principle of neural networkin detail,but you need some background knowledge to deal with convolution network.There is a layered architecture for neural networks.Each layer is composed of some nodes.Each node performs some mathematical operations on an input effectively,and obtains an output by calculation.The input that is provided to any given node node is the weighted sum of the output of the previous layer(and usually the bias item of 1 or 0).The algorithm will learn these weights during training.In order to learn these parameters,the output of running training can be compared with the real value and the weights can be updated by propagating errors in the network.
The Unknown Word
| The First Column | The Second Column |
|---|---|
| horizontal | 垂直的 |
| is reponsible for | 负责 |
| convolution | 卷积 |
| layed architecture | 分层架构 |
| architecture | 架构 |
| mathematical operations | 数学运算 |
| concise | [ken'sais]简明的 |
| by calculation | 通过计算 |
| weights | 权重 |
| propagating | 反向传播['pra:pegeit] |
| an error to be raised | 引发错误 |
| raveling | 拆开[ravling] |
| reshape | 重构 |