当前位置:网站首页>[CV] Wu Enda machine learning course notes | Chapter 8
[CV] Wu Enda machine learning course notes | Chapter 8
2022-07-07 07:49:00 【Fannnnf】
If there is no special explanation in this series of articles , The text explains the picture above the text
machine learning | Coursera
Wu Enda machine learning series _bilibili
Catalog
8 Representation of neural networks
8-1 Nonlinear hypothesis
For an image , If the gray value of each pixel or other feature representation method is taken as a data sample , The data set will be very large , If we use the previous regression algorithm to calculate , There will be a very large computational cost
8-2 Neurons and the brain
8-3 Forward propagation - Model display I
- The figure above refers to a with Sigmoid Artificial neuron of activation function , In terms of neural networks , g ( z ) = 1 1 + e − θ T X g(z)=\frac{1}{1+e^{-θ^TX}} g(z)=1+e−θTX1 It is called activation function
- Neural network refers to a set of Neural Networks , first floor (Layer 1) Called the input layer (Input Layer), The second floor (Layer 2) Called hidden layer (Hidden Layer), The third level (Layer 3) Called output layer (Output Layer)
- use a i ( j ) a_i^{(j)} ai(j) To represent the j j j Layer of the first i i i Activation items of neurons (“activation” of unit i i i in layer j j j), The so-called activation term refers to the value calculated and output by a specific neuron
- use Θ ( j ) \Theta^{(j)} Θ(j) Says from the first j j j Layer to tier j + 1 j+1 j+1 Layer weight matrix ( Parameter matrix ), That's what happened before θ \theta θ matrix ( Previous θ \theta θ It can be called parameter p a r a m e t e r s parameters parameters It can also be called weight w e i g h t s weights weights)
- a 1 ( 2 ) a_1^{(2)} a1(2)、 a 2 ( 2 ) a_2^{(2)} a2(2) and a 3 ( 2 ) a_3^{(2)} a3(2) The calculation formula of has been written in the above figure
- among Θ ( 1 ) \Theta^{(1)} Θ(1) It's a 3 × 4 3×4 3×4 Matrix
- If the neural network is in the j j j Layer has a s j s_j sj A unit , In the j + 1 j+1 j+1 Layer has a s j + 1 s_{j+1} sj+1 A unit , that Θ ( j ) \Theta^{(j)} Θ(j) It's a s j + 1 × ( s j + 1 ) s_{j+1}×(s_j+1) sj+1×(sj+1) Matrix
8-4 Forward propagation - Model display II
Vectorization of forward propagation :
- Put Θ 10 ( 1 ) + Θ 11 ( 1 ) + Θ 12 ( 1 ) + Θ 13 ( 1 ) \Theta^{(1)}_{10}+\Theta^{(1)}_{11}+\Theta^{(1)}_{12}+\Theta^{(1)}_{13} Θ10(1)+Θ11(1)+Θ12(1)+Θ13(1) Expressed as z 1 ( 2 ) z_1^{(2)} z1(2)
- be a 1 ( 2 ) = g ( z 1 ( 2 ) ) a_1^{(2)}=g(z_1^{(2)}) a1(2)=g(z1(2))
- Extend to the whole domain , Activation value of the second layer a ( 2 ) = g ( z ( 2 ) ) a^{(2)}=g(z^{(2)}) a(2)=g(z(2)), among z ( 2 ) = Θ ( 1 ) a ( 1 ) z^{(2)}=\Theta^{(1)}a^{(1)} z(2)=Θ(1)a(1), In addition, you need to add an offset term a 0 ( 2 ) = 1 a^{(2)}_0=1 a0(2)=1
8-5 Examples and understanding I
8-6 Examples and understanding II
The figure above shows the calculation x 1 x_1 x1 XNOR x 2 x_2 x2 The neural network of
From the first floor to the second floor, calculate x 1 x_1 x1 AND x 2 x_2 x2 obtain a 1 ( 2 ) a_1^{(2)} a1(2), Calculation (NOT x 1 x_1 x1) AND (NOT x 2 x_2 x2) obtain a 2 ( 2 ) a_2^{(2)} a2(2)
And then to a 1 ( 2 ) a_1^{(2)} a1(2) and a 2 ( 2 ) a_2^{(2)} a2(2) by x 1 x_1 x1 and x 2 x_2 x2 Calculation x 1 x_1 x1 OR x 2 x_2 x2 The result is x 1 x_1 x1 XNOR x 2 x_2 x2
8-7 Multivariate classification
There are four outputs :pedestrian、car、motorcycle、truck
So there are four output units
Output y ( i ) y^{(i)} y(i) For one 4 D matrix , May be :
[ 1 0 0 0 ] or [ 0 1 0 0 ] or [ 0 0 1 0 ] or [ 0 0 0 1 ] in Of Its in One individual \begin{bmatrix} 1\\ 0\\ 0\\ 0\\ \end{bmatrix} or \begin{bmatrix} 0\\ 1\\ 0\\ 0\\ \end{bmatrix} or \begin{bmatrix} 0\\ 0\\ 1\\ 0\\ \end{bmatrix} or \begin{bmatrix} 0\\ 0\\ 0\\ 1\\ \end{bmatrix} One of them ⎣⎢⎢⎡1000⎦⎥⎥⎤ or ⎣⎢⎢⎡0100⎦⎥⎥⎤ or ⎣⎢⎢⎡0010⎦⎥⎥⎤ or ⎣⎢⎢⎡0001⎦⎥⎥⎤ in Of Its in One individual
respectively pedestrian or car or motorcycle or truck
边栏推荐
- Pytorch parameter initialization
- Tongda injection 0day
- [P2P] local packet capturing
- [guess-ctf2019] fake compressed packets
- nacos
- SQL优化的魅力!从 30248s 到 0.001s
- Asemi rectifier bridge rs210 parameters, rs210 specifications, rs210 package
- Pytest + allure + Jenkins Environment - - achèvement du remplissage de la fosse
- Route jump in wechat applet
- Determining the full type of a variable
猜你喜欢
解决问题:Unable to connect to Redis
自定义类加载器加载网络Class
resource 创建包方式
Common validation comments
buuctf misc USB
[GUET-CTF2019]虚假的压缩包
About some details of final, I have something to say - learn about final CSDN creation clock out from the memory model
图解GPT3的工作原理
[mathematical notes] radian
Robot technology innovation and practice old version outline
随机推荐
[Stanford Jiwang cs144 project] lab4: tcpconnection
After 95, Alibaba P7 published the payroll: it's really fragrant to make up this
After 95, the CV engineer posted the payroll and made up this. It's really fragrant
【斯坦福计网CS144项目】Lab3: TCPSender
misc ez_usb
2022-07-06: will the following go language codes be panic? A: Meeting; B: No. package main import “C“ func main() { var ch chan struct
知识点滴 - 关于苹果认证MFI
Flutter riverpod is comprehensively and deeply analyzed. Why is it officially recommended?
Redis data migration
vus.SSR在asynData函数中请求数据的注意事项
nacos
[SUCTF 2019]Game
UWB learning 1
The annual salary of general test is 15W, and the annual salary of test and development is 30w+. What is the difference between the two?
CentOS7下安装PostgreSQL11数据库
今日现货白银操作建议
2022-07-06:以下go语言代码是否会panic?A:会;B:不会。 package main import “C“ func main() { var ch chan struct
JS plot flot application - simple curve
【p2p】本地抓包
Robot technology innovation and practice old version outline