Machine Learning - Day2 -ML Intro

Edible Instruction

1. Too many formulations and equations, so screen shots of PPT is necessary.
2. My mathematics sucks, so you 'll see many mathematical knowledge interspersed in my all ML learning session.
3. Don’t be afraid of any mathematical symbol, just BAIDU the definition.

What’s Machine Learning?

Machine Learning = Looking for function

What’s the function in math?
In mathematics, a function is a binary relation between two sets that associates to each element of the first set exactly one element of the second set.

The reality of scene:

Different Types Of Functions

Regression

Regression : The function outputs a scalar
Ref: Lean more about regression

Classification

Classification: Given options(classes), the function outputs the correct one.

So Alpha Go is a classification function, 19X19 classes to output a position

Structured Learning

Structured Learning: Create something with structure( image, doctument, video)
Ask machine to create the things.

Example application

Reminder: the function of sample is a conjecture

Estimating the views of Youtube channel tomorrow, base on the history of views

1. Function with unknown parameters
   

2. Define Loss from Training Data (Important!!!)

Loss is a function of parameters L(b,w), and comes from the Training DATA.
Loss: how good a set of values is.
So, if Loss value is large, the parameters are bad.

Suppose L(0.5k, 1)

Label : The true value is label!!!

$\sum$ : Sum things up (called Sigma)
Refer to Article

---

Sum all $e_n$ up

Choose MAE to calculate on the PPT, and MSE on the homework

probability distribution (概率分布)
Cross entropy (交叉熵)

The graph of true data

3. Optimization

arg min f(x) : the value of x when f(x) have the minimum value. arg =
argument, arg min: the argument of the minimum of target function

Gradient Descent:

What’s Gradient Descent?
Please refer to Article and Video

Postulate we only have one unknown parameter

Pick an initial value $W^0$ (Blind guess here)

Differential coefficient(微分), please refer to 知乎acticle
and Differential Equations Solution Guide

Learning rate : $\eta$

Update 𝑤 iterative, and find local minima, can’t find global minima? That’s not true.

Postulate we have two unknown parameters

Graph

Summarize all above:

HOW TO BE BETTER

You need to know:
We’re using the real data(1/1) with function y to calculate the value of (1/2), and iterate to 2/14
We found another law that the views are lower on Friday and Saturday.

Base on the graph, we’ll use a new model.

We’ll learn Linear models on Intro_2.