English Class Review

English Class Review

The Honest Theif

  • meet the theif in the bar, offer him a drink
  • follow everywhere
  • drink all night
  • move to the new house
  • sell his breeches, lost
  • fight
  • go away
  • come back
Read more
Neural Network

Neural Network

The server’s previous technical issues have caused the loss of images and data.

基础的神经单元

参数 功能
$x_i$ 输入的n维度的向量
$w_i$ 加权的系数
$b$ 偏置
$z$ $\Sigma_{i=1}^{n} w_ix_i+b$ 的值
$h(z)$ 经过激活函数得到的一个范围在0-1之间的数
$a$ 作为输入向量x传给下一个神经元
Read more
Matrix Linear Regression

Matrix Linear Regression

人工智能第二讲

矩阵形式的线性回归

image-20200201092909252

x_11表示第一个样本的第一个值

image-20200201093548743

T: 转制, 横竖向量转换

image-20200201094419099

二范数就是各个数的平方和

image-20200201100115985

伪逆矩阵

KNN分类算法

image-20200201103246435

严重依赖于特征

图像数据处理基础

Getting started with Python 2

Getting started with Python 2

For and Else

1
2
3
4
for x in range(6):
print(x)
if x==2: break
else: print("over") # this line will skip if last line exists

Pass

There cannot exist any blank within for and if else. We must use pass to avoid the mistake.

1
2
3
for x in [1,2,3,4]:
pass
print("over")
Linear Regression

Linear Regression

人工智能第一讲

定义

基于已知的函数模型预测未知的

分类

监督学习

  • 需要人工标记
  • 预测、推荐、标注、识别等
  • 回归
  • 分类

无监督学习

  • 聚类
  • 社团划分
  • 生成学习
  • 强化学习: 根据结果回馈优化模型

有监督模型

样本二元组:(x,y)

利用样本求解模型最佳参数

线性回归

image-20200127092012971

image-20200127093925978

  • hyperparameter

image-20200127102104560

Getting started with Python 1

Getting started with Python 1

What’s new for me


  • Do not need to declare variables
  • indentation replaces curly braces to divide code blocks

Grammar


Assigning value to multiple variables

1
x, y, z = 1, 2, "hello"

Define a global variable in local function

1
2
3
def myfunc():
global x
x = "fantastic"

Change the value of a global variable inside the function

1
2
3
4
x = "declan"
def myfunc():
global x
x = "declan and jessica"
Read more

Intro to Neural Network

The server’s previous technical issues have caused the loss of images and data.

线性回归补充

  1. 线性回归可以对样本是非线性的,只要对参数线性

$y=\theta_0+\theta_1x+\theta_2x^2$

  1. 局部加权回归

人工神经网络

相同的词向量表示会非常近

  1. 把字用向量表示

  2. 卷积操作

  3. RNN表示

  4. 平均 pool

  5. 全连接

  6. sigmoid