Upload
ula
View
76
Download
0
Embed Size (px)
DESCRIPTION
the feasibility of machine learning . Component of learning. Formalization Input (输入) :X (customer application) think of it as deed dimension vector Output (输出) :Y(+1,-1) good/bad customer Target Function( 目标函数 ) : f :x → y ideal credit approval formula. Component of learning. - PowerPoint PPT Presentation
Citation preview
Logo
Company Logo
Component of learning
Formalization – Input (输入) :X (customer application) think of it as deed dimension vector
– Output (输出) :Y(+1,-1) good/bad customer
– Target Function( 目标函数 ) : f :x→y ideal credit approval formula
Logo
Company Logo
Component of learning
Formalization – Data (数据) : (), (),…, () historical records
↓ ↓ ↓– Hypothesis( 假设 ) : g :x→y为了得到目标函数的公式
F is unknown G is very much knownactually we created it
Logo
Company Logo
Component of learning
UNKNOWN TARGET FUNCTION f :x→y ↓ ↓
TRAINING EXAMPLES (), (),…, ()
FINAL HYPOTHESIS g
(G hopefully approximates F)
Logo
Company Logo
Component of learning
UNKNOWN TARGET FUNCTION f :x→y ↓ ↓
TRAINING EXAMPLES (), (),…, ()
FINAL HYPOTHESIS g
(G hopefully approximates F)
Logo
Company Logo
Component of learning
TRAINING EXAMPLES (), (),…, ()
LEARNING ALGORITHM →HYPOTHESIS SET( 从现实模型公式中创造公式) (将它们成为假设集)
FINAL HYPOTHESIS
Logo
Company Logo
Component of learning
HYPOTHESIS SET H
从假设集选出一个假设H 衍生出一堆 H’s( 待定函数 )
Logo
Company Logo
Component of learning
UNKNOWN TARGET FUNCTION f :x→y ↓ ↓
TRAINING EXAMPLES (), (),…, ()
FINAL HYPOTHESIS g
(G hopefully approximates F)
Logo
Company Logo
HYPOTHESIS SET
Hypothesis Set – H = {h} g∈ H
Learning Algorithm– Together. they are referred to as the
learning model. a hypothesis set and a learning algorithm
Logo
Company Logo
A simple hypothesis set—’perceptron’
For input X= attributes of a customer– Approve credit if
> threshold
– Deny credit if < threshold
This linear formula h∈ H can be written ash(x) = sign( () – threshold )
Logo
Company Logo
Learning Feasible
Logo
Company Logo
Learning Feasible
A related experiment
P(picking red)=μ P(picking green)=1-μ
μ=probability of red marbles
Logo
Company Logo
Learning Feasible
Pick N marbles independently
The fraction of red marbles in sample = v
Logo
Company Logo
Does v Say anything about μ?
NO!
Sample can be mostly green while bin is mostly red
Logo
Company Logo
Does v Say anything about μ?
Yes
Sample frequency v is close to bin frequency μ
This is called Hoeffding Inequality
Logo
Company Logo
Learning Feasible
Bin – Unknown is a number μ
Learning – Unknown is a function f:x→y
each marble is a point x ∈ X
Logo
Company Logo
Learning Feasible
Bin – Unknown is a number μ
Learning – Unknown is a function f:x→y
each marble is a point x ∈ X
Hypothesis got it right h(x)=f(x)
Logo
Company Logo
Learning Feasible
Bin – Unknown is a number μ
Learning – Unknown is a function f:x→y
each marble is a point x ∈ X
Hypothesis got it wrong h(x)≠f(x)
Logo
Company Logo
Learning Feasible
UNKNOWN TARGET FUNCTION f :x→y ↓ ↓
TRAINING EXAMPLES (), (),…, ()
FINAL HYPOTHESIS g
(G hopefully approximates F)
Logo
Company Logo
Learning Feasible
Probability distribution
P on X