1 CSE 4705 Artificial Intelligence Jinbo Bi Department of Computer Science & Engineering jinbo.

1

CSE 4705Artificial Intelligence

CSE 4705Artificial Intelligence

Jinbo BiDepartment of Computer Science &

Engineeringhttp://www.engr.uconn.edu/~jinbo

The InstructorThe Instructor• Ph.D. in Mathematics• Working experience

• Siemens Medical Solutions• Department of Defense, Bioinformatics• UConn, CSE

• Contact: jinbo@ engr.uconn.edu, 486-1458 (office phone)• Research Interests:

• Machine learning, Computer vision, Bioinformatics• Apply machine learning techniques in bio medical informatics• Help doctors to find better therapy to cure disease

subtyping GWAS

Color of flowers

Cancer, Psychiatric

disorders, …

http://labhealthinfo.uconn.edu/EasyBreathing

3

TodayToday

Organizational details

Purpose of the course

Material coverage

Introduction of AI

4

Course SyllabusCourse Syllabus

Go over syllabus carefully, and keep a copy of it

Course website http://www.engr.uconn.edu/~jinbo/

Spring2015_Artificial_Intelligence.htm

5

Instructor and TAsInstructor and TAs

My office hoursTue 1 – 3pmOffice Rm: ITE Building 233

Two TAsXingyu Cai ([email protected])

office hours Fri 2-3pm, contact him for the place to meetXia Xiao ([email protected])

office hours Fri 2-3pm, ITEB 221

6

Required TextbookRequired Textbook

Attending the lectures is highly encouraged, and lectures highlight some examplesAttending lectures is not a substitute for reading the textRead the text in Chap 1 – 9, because we follow them tightly

7

Optional TextbooksOptional Textbooks

These textbooks cover some of the most popular and fast-growing sub-areas of AI

8

PrerequisitePrerequisite

Good knowledge of programmingData structuresAlgorithm and complexityIntroductory probability and statisticsLogic (discrete math)

9

SlidesSlides

We do not always have slides for later lecture

We use more lecture notes than slides

Slides will be used to demonstrate, and will be available at HuskyCT after the lecture

10

Marking SchemeMarking Scheme

3 HW assignments: 30%

(programming based, and require time to complete)

1 Midterm:30%1 Final Term project: 40%

CurvedCurve is tuned to the final overall distributionNo pre-set passing percentage

11

Grading ArrangementGrading ArrangementXingyu Cai (BECAT A22)Responsible for

HW 1Mid-term examFinal term projects

Xia Xiao (ITEB 221)Responsible for

HW 2HW 3

Please find the right TA for specific questions

12

Questions?Questions?

13

In-Class ParticipationIn-Class Participation

Finding errors in my lecture notes

Answering my questions and asking questions

Come present your progress on term projects

14

Material CoverageMaterial CoverageTwo sets of topics:

classic versus state-of-the-art

Weeks 1 - 9: Intelligent agentsSearching, informed searchingConstraint satisfaction problemsLogical agentsFirst-order logic

Read text chap 1-9 in the required textbook

15

Material CoverageMaterial CoverageTwo sets of topics:

classic versus state-of-the-art

Weeks 10 - 14: Basics in learning (supervised vs. unsupervised learning)Support vector machinesArtificial neural networks

These largely come from the optional textbooks, will give slides to read

16

Course EvaluationCourse EvaluationClassic topics for weeks 1-9

3 HW assignments and 1 mid-term60% of the final grade

Machine learning topics for weeks 10-14

A substantial term project40% of the final grade

17

AssignmentsAssignments

Each will have 4-10 problems from the textbook (not all problems need coding)

Solutions will be published at HuskCT when grades are returned

Each assignment will be given 1-2 weeks to complete, and grades will be returned 1 week after turn in

18

Term ProjectsTerm Projects

Substantial projects require teamwork. Teams of 4-6 students should formed.

Each team needs to present at class their project progress

Each team needs to submit a final report together with necessary codes/results for grading

19

Term ProjectsTerm Projects

Three projects will be designedAll from real-world AI applicationsSpecifically big data applications

1) Drug discovery (computational biology)

2) Disease understanding - Alzheimer’s Disease from images

3) Robotics – learning to move Sarcos robot arm

20

Term ProjectsTerm Projects

Involve learning the background by reading 1-2 papersInvolve programming with any of the following languages/packages

JavaPythonMatlabOr existing ML packages written in these languages

21

Questions?Questions?

22

Why This Course?Why This Course?

A lot to listLet us say“This course will teach us foundational

knowledge of AI, so later we can do research on top of it to 1. build intelligent agents (robots, search engines etc.2. understand human intelligence

3. handle massive BIG DATA … … … “ Exemplar systems …..

23

I want to design a machine that will be proud of me – Danny Hillis

I want to design a machine that will be proud of me – Danny Hillis

24

DARPA Grand Challenge 2005 (driverless car competition)

DARPA Grand Challenge 2005 (driverless car competition)

Stanley won

25

DARPA Urban Challenge 2007 (driverless car competition)

DARPA Urban Challenge 2007 (driverless car competition)

http://archive.darpa.mil/grandchallenge/

26

Significant advances in NLPSignificant advances in NLP

27

28

Search enginesSearch engines

Google search engine

Amazon (online purchase with product recommendation)

Netflix (recommender systems)

29

BIG DATABIG DATA

Big data emerged from biology, engineering, social science, almost everywhere

30

BIG DATABIG DATABig data emerged from biology, engineering, social science, almost every disciplineFor instance, Biology: the big challenges of big data, Nature 498, 255-260, 2013

Need powerful computers to handle data traffic jams

Most importantly, need AI techniques to learn and discover knowledge from data.

31

What is AIWhat is AI

Views of AI fall into four categories

We focus on “acting rationally”

32

Acting humanly (Turing test)Acting humanly (Turing test)

Λ

33

Acting humanly (social robots)Acting humanly (social robots)

MIT Leonardo Robot – isn’t this the cutest robot ever?

34

Acting humanly (social robots)Acting humanly (social robots)

MIT Leonardo Robot – isn’t this the cutest robot ever?

35

Thinking humanly (cognitive modeling)Thinking humanly (cognitive modeling)

36

Thinking rationally (laws of thought)Thinking rationally (laws of thought)

37

Acting rationally (rational agents)Acting rationally (rational agents)

38

Human has much stronger perception than computersHuman has much stronger perception than computers

Can you see a dalmation dog?

39

Survey?Survey?