Overview of Artificial

Intelligence and this Course

Fundamentals of Artificial Intelligence
Instructor: Chenhui Chu
Email: chu@i.kyoto-u.ac.jp

Teaching Assistant: Youyuan Lin

E-mail: youyuan@nlp.ist.i.kyoto-u.ac.jp
Self Introduction
• Experience
• 07/2020-present: program-specific associate professor @ Graduate School of
Informatics, Kyoto University
• 04/2017-06/2020: research assistant professor @ Osaka University
• 04/2015-03/2017: researcher @ Japan Science and Technology Agency
• 04/2014-03/2015: research fellowship for young scientists (DC2) @ JSPS
• 10/2010-03/2015: master & PhD @ Kyoto University

• Research
• Machine translation (JSPS DC2, Chinese-Japanese MT practical application project,
JSPS research activity start-up, Google research scholar, JSPS Scientific Research B)
• Language and vision understanding (JST ACT-I, MSRA CORE, JSPS young scientists)

2
Lecture Attendance
• Time: ⽉2

• Classroom (in principle): 1共02

3
Lecture Handouts
• Lecture handouts will be put in PandA before each lecture

4
Evaluation of this Course
• Class attendance (15%)
• Classroom: KULASIS出席登録システムを利⽤
• 毎回IC カードリーダーに学⽣証を飾ってください
• (For KUINEP students: write down your student ID and name in the class in case
you do not have an IC card)

• Reports (60%)
• 12 Mini reports and exercises

• Final report (25%)

• Solve a real problem in your studies using the models learned in this course

5
AlphaGo
• Beat the human Champions in Go

6
AlphaStar
• Mastering the Real-Time Strategy Game StarCraft II

7
Microsoft Suphx
• The world’s strongest mahjong AI

8
Self-Driving Cars
• We may not need to drive by ourselves in the near future

9
Robots
• Can act similar to humans

10
IBM Watson
• Destroy Humans in Jeopardy

11
Machine Translation
• Achieving Human Parity on Automatic Chinese to English News
Translation

12
Dialogue Systems

13
Very Large Language Models

14
DALL·E: Creating Images from Text

15
ChatGPT

16
AI are Powered by Machine
Learning (Deep Learning)
Paradigm of Machine Learning
• Depends on the different type of training data, can be divided into:

18
Supervised Learning
• Requires lots of labeled samples

19
Supervised Learning Process

Training data yt (GT)

(2) Loss
function
x (input data) (1) Model yp (predicted)

(4) Training (SGD) (3) Gradient

20
Merits of Supervised Learning
• Works amazingly well, when you have data
• Good at: Content understanding, filtering, ranking, translation,
• Services like Facebook, Instagram, Google, Youtube,… are built
around it.

21
Applications of Supervised Learning
• 1. Automated emergency braking system, reduce collisions by 40%
• 2. Tumor detection in mammograms [Wu et al. ArXiv:1903.08297]

22
Semi-Supervised Learning
• Combines a small amount of labeled data with a large amount of
unlabeled data during training

23
Weakly-Supervised Learning
• For each example, requires less annotation

24
Unsupervised Learning
• Without labeled data

25
Reinforcement Learning
• Learn from interaction with environment. Perception->Action-
>Reward
• Give a reward when the action is correct

26
Merits of Reinforcement Learning
• Good at: action generation such as in games and simulations
• AlphaGo: 5 million self-play games.
• StarCraft: AlphaStar 200 years of equivalent real-time play.

27
Self-Supervised Learning
• Predict any part of the input from any other part

28
Focus and Goal of This Course
• This course will attempt to give a sufficiently detailed explanation of
at least a few of the most AI common techniques

• We will focus on supervised machine learning in general and deep

learning in particular

• One goal will be to give practical and working knowledge to students,

so that they can apply what they learned to at least some simple
tasks

29
Schedule
• 1. Overview of AI and this Course (10/2)
• 2. Introduction to Python (10/16)
• 3, 4. Mathematics Concepts I, II (10/23, 10/30)
• 5, 6. Regression I, II (11/6, 11/13)
• 7. Classification (11/20)
• 8. Introduction to Neural Networks (11/27)
• 9. Neural Networks Architecture and Backpropagation (12/4)
• 10. Fully Connected Layers (12/11)
• 11, 12, 13. Computer Vision I, II, III (12/18, 12/25, 1/15)
• 14. Natural Language Processing (1/22)

30
 Overview of This Course
11, 12, 13. Computer Vision 14. Natural language
I, II, III processing
Deep Learning Applications

8. Neural network 9. Architecture and 10. Feedforward

Introduction Backpropagation neural networks
Deep Learning

5. Simple linear 6. Multiple linear

7. Classification
regression regression
Basic Supervised Machine Learning

2. Python 3, 4. Mathematics Concepts I, II

Fundamental of Machine Learning 31

Python

32
Mathematics Concepts
• Given a function of several variables, find the value for which it is
minimum
• a.k.a “multivariate function”

𝑓: ℝ! → ℝ
𝑓(𝑥, 𝑦) = (𝑥 + 𝑦)! + 1
argmin𝑓(𝑥, 𝑦)
",$

33
Regression
• The linear regression consists in finding the good values for the
parameters

34
 Classification

Inputs:

Multiclass Classifier
Which animal? (cat, dog, lion, bird, fish, human, other)
(7 classes)

Predicted Class: Cat Dog Bird Cat

35
 Neural Networks
Layer 1 Layer 2 Layer 3 Layer 4 Layer 5

Neuron

Neuron

Age Neuron Neuron

Neuron

Neuron Neuron Neuron Vote

Income Neuron Neuron

Neuron

Neuron

36
 Computer Vision
• Having computer understand images the way human do
Object detection Image Recognition
CAT

Emotion Recognition

SAD
Image: http://vision.seecs.edu.pk/ats/

Yes
Is phone
“FaceID” owner?

No
37
 Natural Language Processing
• Having computer understand text the way human do

Russian defense minister Ivanov called Summarization

Russia calls for a joint front
Sunday for the creation of a joint front for
combating global terrorism against terrorism

Imdb comment This user thinks the

movie is bad or good?
Since Disney is incompetent of coming up with new Sentiment analysis
ideas, and must resort to using older stories they did
years ago, they certainly better live up to what Bad
they're doing. This does not happen with Mary
Poppins Returns.
38
Summary
• Paradigm of machine learning
• Supervised learning
• Semi-supervised learning
• Unsupervised learning
• Reinforcement learning

• Overview of this course

• Fundamental of Machine Learning
• Basic Supervised Machine Learning
• Deep Learning
• Deep Learning Applications

39