Deep Learning Tutorial
李宏毅
Hung-yi Lee
Deep learning
attracts lots of attention.
• I believe you have seen lots of exciting
results before.
Deep learning trends
at Google. Source:
SIGMOD/Jeff Dean
This talk focuses on the basic techniques.
Outline
Lecture I: Introduction of Deep Learning
Lecture II: Tips for Training Deep Neural Network
Lecture III: Variants of Neural Network
Lecture IV: Next Wave
Lecture I:
Introduction of
Deep Learning
Outline of Lecture I
Introduction of Deep Learning
Why Deep?
Let’s start with general
machine learning.
“Hello World” for Deep Learning
Machine Learning
≈ Looking for a Function
• Speech Recognition
“How are you”
f
• Image Recognition
• Playing Go
f
f
• Dialogue System f
“Cat”
“Hi”
(what the user said)
“5-5” (next move)
“Hello”
(system response)
Framework
Image Recognition:
f
“cat”
A set of
function
Model
f1 , f2
f1
f1
“cat”
“dog”
f2
f2
“money”
“snake”
Framework
Image Recognition:
f
“cat”
A set of Model
f 1 , f 2
function
Better!
Goodness of
function f
Training
Data
Supervised Learning
function input:
function output: “monkey”
“cat”
“dog”