Teaching Portfolio

 

KAN Min-Yen
 Department of Computer Science
National University of Singapore

9 September 2008
Foreword

This document may be longer than the usual teaching portfolio, but I hope you will be able to navigate it with the help of the indexing.  Where important to distinguish between the original material and my annotations, I have included my annotation in a box such as this one.  Please read these comments, as they will illustrate my reasoning for including this in the teaching portfolio.  The exception is when I have included presentation slides, in which case my annotation will appear as normal text.

 

Kan Min-Yen

9 September 2008

Table of Contents

 

Philosophy........................................................................................................................ 3

Treat as a peer............................................................................................................. 3

Engage their thinking and creativity......................................................................... 3

Ask them to help you................................................................................................... 4

 

Sample Materials from Module Folders....................................................................... 6

Samples from CS 3243 – Foundations of Artificial Intelligence.......................... 7

Samples from CS 6210 / 5244 – Digital Libraries............................................... 24

 

Selected Student Feedback........................................................................................ 33

 

Appendix A – Emails on slide development............................................................. 36

Appendix B – IOI / ACM Training Slides.................................................................... 38

Appendix C – Teaching Practicum report excerpt................................................... 39

Appendix D – Beginning research presentation excerpt....................................... 42

Appendix E – Graduate Research Course Joint Poster Seminar........................ 43

Appendix F – Examples of Discussion Questions................................................... 44

 

The corresponding, complete student feedback for all of the modules I have taught can be found in the section following this one (Student Feedback).  The modules folders (Mod. Folder CS5244, Mod. Folder CS5246, Mod. Folder CS3243) found in the subsequent tabs give a complete inventory of:

 

- Web pages (which includes syllabi and course objectives

- Presentation slides

- Sample final exam


Philosophy

This is an expanded version of my teaching philosophy statement found in the main dossier.  It just contains additional evidence and claims to support my teaching beliefs.

 

While, by any rights, I am fairly new to formal teaching, I have been tutoring friends and family throughout my life.  As tutoring is defined as giving individual instruction, it may seem that best practices in tutoring may not extend well to teaching.  And one would be right; I would say that many aspects of teaching many cannot be done in the way of tutoring one.  However, the principles that I used to effectively tutor can be adapted to produce effective teaching. 

 

My teaching philosophy is to view teaching as tutoring many simultaneously.

 

What does this mean?  At first, when I developed this idea, it was a mental kludge to protect myself from being nervous from being watched and assessed by so many students.  I felt that it was easier to think of the students as a single student, whose opinion, understanding and interest level was a product of all of the students of the class.  This let me interact more naturally with the students and released me from the lock of nervousness.  

 

While I don’t have as many problems with nervousness as I did previously, the concept of tutoring still remains a powerful one that has many strong analogies that I believe is true of great teaching as well.  Three components of effective tutoring sum up my approach rather nicely.

Treat as a peer

A tutor treats their student as a peer, not as an individual of lower standing.  This means that as tutors and teachers, we extend to our students the rights and respect that we would demand of and give to our peers.  Examples of this start with basic courtesy: remembering their names when possible and encouraging them to address me by first name.

 

“… and he even knows every student by name (17 students)”

[ from my last year’s annual review ]

 

Treating students as peers also means that they should know right from wrong.  I make it very clear in my upper-class modules that I trust their judgment.  I pointedly encourage students to openly discuss any homework problem and projects, but ask that they do not take away any form of notes (electronic or otherwise).  In my classes, this has come in the form of the Gilligan’s Island Rule (as can be seen in my CS 3243 module folder, page 11), which I discuss and append to the course website every semester.  This level of trust is reciprocated back, which I feel helps to bond my students to me.  Another example is in the way I deal with student emails.  When a student email touches a subject that I think may be of interest his fellow students, I will anonymize the email and post the email and my reply to the IVLE forum.  I tell the students about this policy in the beginning of the semester, and I take special care to avoid posting any personal information.  I feel that this encourages students to ask more questions, as they see that some of their concerns have been voiced by others.  As a student put it nicely:

 

“I like the way he 'broadcasts' emails sent to him regarding the module in IVLE. Other lecturers don't do that. I feel that his style will benefit most people.”
[ CS 3243 feedback ]

 

One may believe that remembering student’s names and interests lies far from the tenets of good teaching.  I have found that the knowing the student better often can yield better learning outcomes, as students are tuned in for lectures and expect to be engaged as a peer, rather than lectured at.

Engage their thinking and creativity

When an instructor steps off of the podium and enters the class as a peer, the requirements of him change as well.  When tutoring students one on one, I seek to engage my peer and try to get them to think of issues in the larger context as well as relevant examples.  After all, when tutoring, we don’t want to help the student solve a particular problem, but rather to instill in them how to learn to solve problems independently of us.  To do this, we must continually search for relevant materials to share with our students.  An example of this is in the introduction of a new lecture on new media in my Digital Libraries class.  Blogging, instant messaging and SMSing have become new outlets for communication and knowledge exchange.  These technologies are used by virtually every student but by reflecting upon it in the course setting allows students to think about the technology from the fresh pespective of the evolution of knowledge exchange, and not just from a user’s perpective. 

 

Another example is in my use of choice of second homework assignment in my Artificial Intelligence class.  Students have different interests and A.I. covers a wide swath of material.  Instead of constructing a single homework assignment, I constructed three different ones.  Students assigned themselves to homework assignments that involved either vision, robotics or language processing.  The outcome of this extra effort paid off in the way students perceived and reacted to the assignments:

 

He is one of the nicest lecturers I've met so far in my 2 years university learning period. I nomite him for his professionalism and patience to us students. Another remarkable point is that he really THINKS how this module can improve, we can see that through the way he designs projects and so on. Thanks!
[ CS 3243 feedback, emphasis due to the student ]

 

Is creating choice valuable to students?  The answer is yes.  My teaching practicum project focused on this question and I think it is relevant to the teaching portfolio, so I have included it as Appendix C.

 

In tutoring, we are taught to guide the student rather than solve the problems for them.  This is true of the most valued teaching.  To this extent, I like to guide my HYP and UROP students in beginning steps in their research, as shown in Appendix D, in which I deconstruct the qualities of a good undergraduate project.  I also use this outline that I give to my new students for guiding my students in my postgraduate course.

Ask them to help you

After all, your students are your peers, so they can help improve our teaching.  This has taken two forms in for me thus far: the mid-term instructor evaluation and the final evaluation with my project students. 

 

As prescribed by CDTL, the mid-term evaluation is really a great device to get some constructive criticism on teaching skills.  In past semesters, I had implemented this on my own (see the module folder for CS 6210), but in the past semester, CITA initiated its own midterm feedback.  I opted at that time to use their midterm feedback form rather than mine own.  In hindsight, this was a mistake, as the CITA form encouraged quantitative evaluation rather than constructive qualitative feedback. 

 

Feedback from a previous semester showed me that students coming to class late (sometimes due to circumstances beyond their control) wanted to have a webcast to view the key points they missed.  As such, I have mandated that all of my classes be webcasted, so that students can have control over their learning process.  This has been effective with graduate students as well, as many taking my current module are part-time, off-campus students.  While many do take the pains to come into class, I can understand the cost of doing so, and would like to make attending class suitable for every and any student who wishes to learn.

 

With my research project students, I am lucky to interact with them over a longer period of time and to do it one-on-one.  I conduct a final exit interview after UROP and HYP deadlines are over to get their frank feedback from them about their what they think of their work and my method of advisement.  Building a good rapport and sense of openness with my students is crucial for any supervisor.  My undergraduate project students have told me informally that they feel that they get a lot of attention from me, as a meet with each of them on a biweekly basis for an hour each.  While this consumes a large portion of my time (18 undergraduate students = 9 hours per week), my expectations of them are high and they have risen to my challenge. 

 

Min-Yen Kan

9 September 2008

 

 

 

Sample Materials from Module Folders

Rather than present all of the course material, I have extracted a few selections from components from each module and compiled some notes to guide your assessment.

 

You can find all content of the module in IVLE and on the class home pages.  I do not remove past year’s homepages, rather I use a different set of web pages each year to indicate the difference in class. 

 

If you do not have the time to review the entire module folder, I suggest that you read the highlights given in the summaries.

Samples from CS 3243 – Foundations of Artificial Intelligence.......................... 7

Aims and Objectives................................................................................................ 7

Syllabus..................................................................................................................... 8

Teaching-Learning Materials.............................................................................. 10

Continual and Final Assessment........................................................................ 11

Homework #1 - Pipe dream............................................................................. 11

Implementation.............................................................................................. 13

Getting a feel for the assignment................................................................ 13

Compiling and submitting the assignment............................................... 14

HW 1 - Excerpt from the high score board................................................ 15

Scores per agent........................................................................................... 15

Comments from Students on HW 1............................................................ 16

Homework #2 - Implementing a Robotic Arm using LEGO Mindstorms. 17

Background.................................................................................................... 17

Task Specification......................................................................................... 17

Software Design and Implementation....................................................... 18

Hardware Design and Implementation..................................................... 18

Testing the Robotic Arm............................................................................... 19

(Bonus part): Online Planning..................................................................... 19

Resources....................................................................................................... 20

Pictures from their work................................................................................ 20

Final Exam Questions....................................................................................... 22

Summary................................................................................................................. 23

 

Samples from CS 6210 / 5244 – Digital Libraries............................................... 24

Course Description............................................................................................... 24

Aims and Objectives............................................................................................. 24

Modes of Learning................................................................................................ 24

Textbook and Required Readings..................................................................... 25

Syllabus................................................................................................................... 26

Teaching-Learning Materials.............................................................................. 27

Continual and Final Assessment........................................................................ 30

Homework 2 - Authorship Detection.............................................................. 30

Details.................................................................................................................. 30

What to turn in.................................................................................................... 30

Grading scheme................................................................................................ 31

Summary......................................................................................................................... 32


Samples from CS 3243 – Foundations of Artificial Intelligence

In A.I., I strove to add value to an existing course.  As mentioned, the curriculum was revamped to center around two interesting homework assignments.  The second assignment, in particular, revolved around a choice of three different assignments in different areas.  We (myself and the TA) had to offer some pre-recorded lectures, see the Syllabus (page 12) to let students pursue their own interests. 

Aims and Objectives

I use IVLE for temporally sensitive things, but the majority of the coursework material I place directly on the web, allowing students from other universities to access this information.  This has paid off tremendously, as I have received comments from both instructors and students alike asking for materials.

We will be using the Integrated Virtual Learning Environment (IVLE) for forum discussions, announcements, homework submissions, and other temporally-sensitive materials. Basic course administration, lecture and tutorial notes will be available on this publicly-accessible webpage.

Description (from the course bulletin): The module introduces the basic concepts in search and knowledge representation as well as to a number of sub-areas of artificial intelligence. It focuses on covering the essential concepts in AI. The module covers Turing test, blind search, iterative deepening, production systems, heuristic search, A* algorithm, minimax and alpha-beta procedures, predicate and first-order logic, resolution refutation, non-monotonic reasoning, assumption-based truth maintenance systems, inheritance hierarchies, the frame problem, certainly factors, Bayes' rule, frames and semantic nets, planning, learning, natural language, vision, and expert systems and LISP.

Prerequisites: CS1102 Data Structures and Algorithms, and either CS1231 or CS1231S Discrete Structures
Modular credits: 4
Workload: 2-1-0-3-3

Our textbook will be the same textbook as last semester. You can buy it from your seniors for it or from the NUS Co-op. Make sure you get the second edition textbook.

Copies are also on reserve (RBR) at both the Science Library and the Central Library.

Note to NUS-external visitors: Welcome! If you're a fellow AIMA instructor looking for MS Powerpoint versions of the AIMA slides, you can see the syllabus menu item on the left for a preview, please contact me for the original .zip file.

Here, I want to highlight my trust in my students.  In this section, you can read my discussion on the academic honesty policy and read my policy of using the Gilligan’s Island Rule to encourage collaboration to allow students to self-police themselves about the appropriate level of collaboration.

Academic Honesty Policy

(Updated Thu Jul 24 14:15:05 GMT-8 2003)

Collaboration is a very good thing. Students are encouraged to work together and to teach each other. The extra-credit programming homework will require a team effort in which everyone is expected to contribute.

On the other hand, cheating is considered a very serious offense. Please don't do it! Concern about cheating creates an unpleasant environment for everyone.

So how do you draw the line between collaboration and cheating? Here's a reasonable set of ground-rules. Failure to understand and follow these rules will constitute cheating, and will be dealt with as per University guidelines. We will be policing the policy vigorously.

You should already be familiar with the University's academic code. If you haven't yet, read it now.

This section on academic honesty is adapted from Surendar Chandra's course at the University of Georgia, who in turn acknowledges Prof. Carla Ellis and Prof. Amin Vahdat at Duke University for his policy formulation. The Gilligan's Island rule origin is uncertain, but at least can be traced back to Prof. Dymond at York University's use of it in 1984.

Syllabus

Here you can see the added, pre-recorded lectures that we had to do to accommodate the choice of homework assignments we offered.  It is highlighted in red.

(Last updated on: Tue Mar 23 19:37:10 GMT-8 2004 ) The exam begins at 2:30 pm on the 23rd of April, not previously announced at 2:00 pm. Thanks!

Your teaching staff will try their best to provide you with detailed assessment marks within three weeks of the submission due date. You are welcome to debate/argue for any extra points that you feel that you deserve within one week after the initial grades have been released.

Any questions about this information should be directed to the general forum on IVLE. The lecture notes have been adapted from earlier notes by Prof. Ng Hwee Tou and Prof. Tan Chew Lim, which in turn have been adapted from the original slides from Stuart Russell and Peter Norvig. Our teaching assistant, Mr. Huang Weihua, contributed the slides for the optional lectures on vision and robotics. There's also a good set of complementary notes by the MIT OpenCourseWork project. You might want to consult and/or print them out. Thanks to those of you took the time during the midterm feedback to point this out!

Exam Date: 23 April 2004 (Afternoon)

Date

Description

Deadlines

S1: 7 Jan

Introduction to AI and Agents

Orientation (v 1.0) [ .pdf ] [ .htm ]
Introduction (v 1.0, Chapter 1) [ .pdf ] [ .htm ]
Intelligent Agents (v 1.0, Chapter 2) [ .pdf ] [ .htm ]

- IVLE survey

S2: 14 Jan

Blind Search

Solving problems by searching (v 1.0, Chapter 3) [ .pdf ] [ .htm ]
Bi-directional search addendum, 2 slides (v 1.0, Chapter 3) [ .pdf ] [ .htm ]

 

S3: 21 Jan

Informed Search

Informed Search and Exploration (v 1.0, Chapter 4) [ .pdf ] [ .htm ]
Online search and homework 1 addendum, 5 slides (v 1.0, Chapter 4) [ .pdf ] [ .htm ]

- Tutorials begin
- Homework 1: Pipe dream game playing

S4: 28 Jan

Game Playing

Adversarial Search (v 1.1, Chapter 6) [ .pdf ] [ .htm ]

 

S5: 4 Feb

Constraint Satisfaction

Constraint Satisfaction Problems (v 1.1, Chapter 5) [ .pdf ] [ .htm ]

Homework 1 due

S6: 11 Feb

Midterm
45-minute lecture follows midterm: Introduction to Advanced AI Topics

Introduction to Advanced Topics (see misc for videos) [ .pdf ] [ .htm ]
Algorithms Addendum for Chapters 2-4,6 (v 1.1) [ .pdf only ]

Computer Vision (v 1.0, Chapter 24) [ .pdf ] [ .htm ]
Natural Language Processing (v 1.0, Chapter 22) [ .pdf ] [ .htm ]
Robotics (v 1.1, Chapter 25) [ .pdf ] [ .htm ]

- Midterm test: Closed Book
- <Pick up an advanced topics lecture on your own between S7 and S9*>
- Self-selection of teams, and assignment of homework #2 preferences

S7: 18 Feb

Logical Agents and Propositional Logic

Logical Agents (v 1.2, Chapter 7) [ .pdf ] [ .htm ]

Team Homework 2 released.

S8: 25 Feb

Logical Agents and Propositional Logic (cont'd, Addendum to previous week's notes) [ .pdf ] [ .htm ]

 

 

S9: 3 Mar

First Order Logic

First-Order Logic (v 1.2, corrected slide 16, Chapter 8) [ .pdf ] [ .htm ]
Inference in First-Order Logic (v 1.1, Chapter 9) [ .pdf ] [ .htm ]

 

S10: 10 Mar

Uncertainty

Uncertainty (v 1.1, Chapter 13) [ .pdf ] [ .htm ]
Probabilistic Reasoning (v 1.1, Chapter 14) [ .pdf ] [ .htm ]

 

S11: 17 Mar

Learning

Learning from Observations (v 1.1, Chapter 18) [ .pdf ] [ .htm ]

Team Homework 2 due (extended by one week as of Mon Feb 16 17:50:22 GMT-8 2004)

S12: 24 Mar

Exam Revision (Check out misc for past exams)

Revision (v 1.0) [ .pdf ] [ .htm ]

Last Tutorial

23 Apr
(Friday)

Final Exam (Afternoon / Closed book; one A4 double-sided sized sheet allowed with notes)
2:30 pm at the PGP residences multi-purpose hall

Min-Yen Kan <kanmy@comp.nus.edu.sg> Created on: Mon Dec 1 19:36:22 2003 | Version: 1.0 | Last modified: Sat Mar 27 16:17:08 2004


Teaching-Learning Materials

Most of the lecture material in the course was from previous semesters work by Ng Hwee Tou and Tan Chew Lim.  Therefore I have not shown instances of the slides used for the course.  My contribution here was to translate the slides for the course from SliTeX to Powerpoint.  This may not sound like much work, but it was a sorely needed addition to the teaching materials for the community.  I refer you to Appendix A for my contribution.

Continual and Final Assessment

I’ve included the first assignment on game playing, and one of the three choices of the second assignment (Robotics: construct a robotic arm).  I’ve also included portions of my final exam to illustrate the types of questions that I offered.  My former advisor also asked for the homeworks to see whether she could use them in her class (see email on page 27)

Homework #1 - Pipe dream

With HW 1, the students had to implement a game player.  Some students told me it took them many hours because they were motivated to do it.  I’ve included some of the high scores board to showcase their efforts.

(Adapted from the original Japanese arcade game in 1990)

For this homework, you will implement a game playing agent to play the solitaire game, Pipe Dream. In Pipe Dream, the goal is to lay pipe to divert ooze from a gushing sewer main break into a system of pipes to keep it from leaking out as long as possible. Your score is the final determiner of whether your agent does well or not.

The board is a 7x7 grid in which a pipe tile can be placed in a square. The pipe tiles can be placed anywhere on the grid and come in seven different configurations: a vertical pipe, a horizontal pipe, a cross (a vertical and horizontal pipe stacked together) and four elbow pipes. Pipe tiles are provided one at a time, in some order. In addition to knowing the current piece to be placed, the agent has access to the queue of the next 4 tiles to be placed. The source of the ooze originates from the center tile, going towards the top.

The initial layout at time 0.

At time t=0 through t=6, no ooze emerges from the center source tile. Subsequently, after 3 units of time, the ooze will go through a single pipe tile. Ooze can be directed through cross pipe in both directions. The grid's edge is the boundary; any pipe ends facing edges can not be connected to additional pipe.

The ooze is starting to flow, at time 6. If there is no suitable tile above the source tile by t=9, the game is over.

At each time unit, the agent can do one of the following actions, all of which take a unit of time to perform:

·     Place the current tile at a empty x,y coordinate

·     Destroy an occupied tile at a filled x,y coordinate (only possible if ooze has not started to flow through the tile yet)

·     Do nothing

At time 18, this agent has laid some pipe, and waited at some time units.

Note: You cannot directly discard a pipe tile; you must place in on an available spot on the grid and then destroy it. If your module specifies to do an action which is invalid, a unit time will also elapse.

The game ends when the ooze spills out of the pipe (because your agent didn't lay an appropriate at the outflow coordinate, or because the flow ran off the edge of the grid, or because the entire grid is filled with pipe. The agent's goal is the maximize its final score (the score is the agent's utility).

Your agent's score can be calculated by the following:

·     Each tile where the ooze flows through: +2

·     Bonus for each cross tile where the ooze flows through both ways: +5 (a total of (2x2)+5 = 9)

·     Each destroyed tile: -3

Note that tiles laid but un-oozed through do not factor into the agent's score whatsoever.

Implementation

Your agent will play this game by way of a single Java class which will extend the abstract class, Player. Most significantly, the Player class has the public method, play(). play() takes a single argument Driver d, an instance of the Driver class, which will enable you to get the state of the game components: the queue, the board and information about the next ooze.

play() will return an Action object, that encodes the agent's action and, when applicable, its coordinates. The class Action, has the following methods:

·     setAction (char action): sets the action of the agent for this unit t to one of the three actions: 'p' place current tile, 'd' destroy tile, 'w' wait. The default value of action is 'w' wait.

·     setX (int x): sets the x coordinate for the action's location. Valid values will be between 0 and 6.

·     setY (int y): sets the y coordinate for the action's location. Valid values will be between 0 and 6.

·     constructor Action (char action, int x, int y): the obvious constructor. Will take the values provided and install it in the appropriate fields.

Note that the valid values for the X and Y coordinates will not check whether the action would actually be valid. It is your responsibility to generate an agent that constructs a valid Action object.

Queue pieces are represented by tiles which take on character values: one of the following characters seven characters: '+' <cross piece>, '|' <vertical pipe>, '-' <horizontal pipe>, or 'q' <tile with left and up outlets>, 'e' <pipe with up and right outlets>,'c' <right and down outlets>, or 'z' <down and left outlets> (it may be helpful to visualize this on a QWERTY keyboard with 's' in the center).

The Board will have two additional types of pieces that can appear: ' ' <empty space>, '\'' <starting tile, with ooze coming out northwards>.

For convenience, in implementing the agent class, you may also call the following accessor methods to get the state of the game, through the Driver class instance:

·     int getTimeToNextOoze (): returns the number of turns left before the ooze moves forward another square.

·     int getTime (): returns the number of turns elapsed since the start of the game

·     int getScore (): returns the agent's score so far

·     int getNextOozeSpotX (): returns the X coordinate of the next spot where the ooze will move to.

·     int getNextOozeSpotY (): returns the Y coordinate of the next spot where the ooze will move to.

·     char[] getQueue (): returns a 5 element array containing the pieces in the queue to be placed. The 0th indexed piece is the current piece to be placed.

·     char[][] getMap (): returns a 7 by 7 element array containing the pieces on the map.

Additional accessor and print methods are available to you in this assignment. The full documentation for the java package is available. Examine the contents carefully. Also available to you is a sample class, MyPlayer.java, that also shows how to compile your agent into an actual game execution.

Getting a feel for the assignment

Well, by all means you have to play the game a bit to get an idea of what type of heuristics and search might be useful. You can try my implementation of the Strategic Pipe Dreams Framework. It'll probably one of the few times you can tell your parents that you must play games in order to do your homework!

Compiling and submitting the assignment

When the pipedream package is released, you'll have to do the following things to compile and run the package.

To compile, you'll need the pipedream.jar (Version 1.1, updated Mon Jan 26 18:38:21 GMT-8 2004) and the directory where you have your implementation of the MyPlayer class in your classpath. Assuming MyPlayer.java is in the current directory:

javac -classpath .:pipedream.jar MyPlayer.java

To execute the game you have to do something similar:

java -cp .:pipedream.jar MyPlayer

You should submit your working version of the agent by the midnight on the due date, 4 Feb. The IVLE workbin will accept late submissions. Late submissions are subject the the late submission policy (read carefully). Your class should be named MyPlayer and should present in a file named MyPlayer.java. Your work should be done independently; see our classes' section on Academic Honesty Policy if you have questions about what that entails.


HW 1 - Excerpt from the high score board

I allowed them to identify their submission with a tagline and then posted the test results to the public.  Here are some.  Some of their taglines are quite humorous.

N.B. - This is the default score as retrieved by the testing harness. The game players are run over 200 random test cases. The scores here do not account for time and computation limits so your actual mileage (and grade) may vary.

Read in 158 agent(s) scores and 156 identities.

Jump to: [ agent averages ] [ best per game ] [ best average over all agents ]

Scores per agent

Average Score

Agent Tagline

59.68

Game playing is fun! But making agents to play for me is not fun! :P

57.38

My agent is smart...

54.68

Game Programming is FUN!

53.38

When testing, please initialized a new Player to play each game. you are really appreciated

47.50

ph34r M01 l33t 5k1||z!!!

47.31

If at first you don't succeed, you are running about average.

44.75

This assignment is fun yoo!

43.79

PipeDream agent using searching and heuristics

42.00

This is the PipeDream Programme written by NAME, MATRIC, SoC Year 1.

40.07

Game playing is fun! Making agents to play for me is even more fun!

39.34

Game playing is painful! Making agents to play is even more miserable!

38.23

Game playing is fun! Making agents to play for me is even more fun!

38.16

Making agents to play for me is *_*? >_
 
Comments from Students on HW 1

 

From:

GAN HIONG YONG

Date:

06/02/2004 10:54:00 PM

 Heading:

Default Heading

Topic:

OMG the scoreboard

 

60+???

  

50+???

  

  

(maybe the lecturer's and tutor's agents join in the fun? lol)

From:

NGUYEN THUY TRAM

Date:

05/02/2004 12:02:00 AM

Topic:

quite interesting... ^^

 

Yo...haha.. homework 1 is over.. i have something to submit ..i thought i couldnt implement it, but i could...not bad

Arghh..but have to say good luck to me haha..

And good luck to ya guys toooooo..

 

Homework #2 - Implementing a Robotic Arm using LEGO Mindstorms

In Homework #2 (robotics variant), they had to implement a robotic arm.  I’ve included some pictures of the resulting work by the students.

You should submit your working version of the assignment by the midnight on the due date. The IVLE workbin will accept late submissions. Late submissions are subject the the late submission policy (read carefully). Your submission should contain a plain .txt file called students.txt which should have the same format as given in the hyperlink (in fact, why don't you save the linked file now and adapt it before you submit it?). Make sure your files are in plain ASCII text format -- MS Word files or other word processing files will NOT be entertained. Your team's work should be done independently of other groups; see our classes' section on Academic Honesty Policy if you have questions about what that entails.

This homework specification was written by Huang Weihua. Please address your questions, concerns and answers to him.

Background

The LEGO Mindstorms product is a robot-kit-in-a-box. It consists of a computer module called the RCX (Robot Command System), and an inventory of many TECHNICS parts. Furthermore, it provides sensors and electric motors that are essential to a physical agent. The following sensors are provided:

Using the LEGO Mindstorms kit, various robots can easily be created, including mobile robots (with either legs or wheels), robotic arms etc. In this assignment, we are going to have a hands-on experience by actually implementing and programming a robotic arm using the LEGO Mindstorms. Some examples of such robotic arms are shown below:

Examples of Lego Robotic Arms

Task Specification

The basic function of a robotic arm is to stretch itself to certain location in the environment, and then perform some actions such as grabbing an object or releasing an object. If a robotic arm works in a blocks world (see textbook pg. 446), then its task is to move the blocks in a planned sequence and reach the goal state.

To simplify our task, we may assume that there are only three locations A, B and C that can be used to put blocks. The blocks in the initial state are put in one of the locations, as shown in Figure 1(a). In the goal state, the blocks are all shifted to another location, and the order among the blocks may change. The first task is to formulate the state representation by determining how you are going to reflect the locations of the blocks and the order among the blocks. Using the state representation, how do you represent the state in Figure 1(a)? How do you represent the state in Figure 1(e)?

Software Design and Implementation

The actions of a robotic arm are controlled by a program. The program generates plans based on the state information and determines the type of an action as well as the parameters required for performing that action. If a continuous planning is adopted, then the planning is based on the current state and one action is returned at a time. A simpler approach is to create a complete plan consisting of a series of actions that are to be carried out to reach the goal state. Using Figure 1 as an example, based on the initial state in 1(a) and the goal state in 1(e), the series of actions are: 

Move(block1, locationA), Move(block2, locationA),
Move(block3, locationC), Move(block2, locationC),
Move(block1, locationC)

Continuous planning requires object recognition that is not achievable using the sensors provided by LEGO Mindstorms kit. Thus you are only required to perform one-time planning at the beginning to determine the series of actions to be carried out by your robotic arm. Now you should come out with your algorithm to generate the action sequence based on the state representation you have come out previously.

Although LEGO Mindstorms Kit already includes a programming software with graphical interface, it is not convenient to write complicated codes that achieves tasks like planning. Fortunately there is a platform called LEJOS which allows us to write JAVA programs and then convert into instructions that are understandable by the RCX. You can find the technical details of the LEJOS in the resources listed in the last section. Now write a program to give your robotic arm the software support. In your program, you should perform two tasks:

1.     To come up with a planner that generates a series of actions to be performed.

2.     To translate the actions into executable instructions and feed them into the RCX.

Note that the initial state and the goal state will be given on the spot, thus hard coding your robot to handle just a specific situation is a bad choice.

Hardware Design and Implementation

A simple robotic arm has 3 effective degrees of freedom (DOF): one for turning the arm to left and right, one for raising and lowering the arm, and one for grabbing and releasing. If each of these 3 DOFs is taken care by an electric motor, then we need a total of 3 motors. The actions that can be performed are then: 

Grab(Object o);
Release(Object o);
RotateLeft(time t);
RotateRight(time t);
RaiseArm(time t);
LowerArm(time t);

Note that the major movement of a robotic arm is rotation in our application. Suppose the power of the electric motor is fixed during the process, then the rotational angle is proportional to the execution time t. By specifying the time t, the movement of the components of the arm can be controlled.

You should carefully design your robotic arm. A suggested way is to construct the two major parts separately.

When you are constructing the two parts you should consider the following issue:

There is an example of robotic arm in the CD-ROM attached with each Mindstorms Kit, you may want to use it as the starting point for your robotic arm, but you should make improvements so that the arm is able to handle the current problem efficiently.

Testing the Robotic Arm

After you have implemented the robotic arm, you may first let it perform the basic actions, such as rotating for different angles etc. After you are sure that your robot is ready, you may define some initial states and goal states, run your program to determine the action series and then feed them into the robot to carry them out.

For testing purpose, you need to construct a physical environment to simulate the "blocks world". If your robotic arm has difficulties grabbing a block, you may consider other alternatives such as a ball or something with irregular shapes that are easier to grab. Essentially, the ordering among the objects should be reflected.

During the final submission, you should include your environment settings so that we are able to test your robotic arm. We may also consider to conduct demo sessions so that you can show how your robotic arm works by yourself.

(Bonus part): Online Planning

Note that in the basic part your task is to simply perform off-line planning based on the given initial state and known goal state. When planning is completed, a series of actions are generated and executed with no further changes. Off-line planning does not involve perception (in this case the light sensors), and it expects all the actions to be performed successfully. However this expectation is too good to be true in most situations and a sensorless planner does not have ability to get out from an unexpected state resulted from a bad action.

In the real world, we have to deal with the case where certain action fails to provide an expected result. Planning methods that handles indeterminacy is discussed in the textbook page 431. The basic idea is to equip the agent with sensors so that it is able to detect any problem during the process of action execution and then manage to recover from it. "Execution monitoring and replanning" is the most appropriate method to be used here.

To apply this approach to your robotic arm, you need to add two features:

To illustrate the idea of localization, let's consider the example given in Figure 1 in the description of the basic tasks. Instead of telling the robotic arm the initial state is 1(a), we may let the arm determine the initial state by itself. When the arm is lowered to some level, the light sensor can tell the color of the block 1, so it knows block 1 is at the top. When the arm moves block 1 to another location and comes back, it sees block 2. Now the initial state is known, since there is no other possible location for block 3 except under block 2.

To show how replanning can be done, let's consider the move from 1(d) to 1(e). The original plan is to move block 1 to location B, followed by moving block 2 to location C, and then move block 1 to location C. Imagine when moving block 2, it drops on block 1 at location B. By execution monitoring, the robotic arm will go to location C and check the color of the top block, which is block 3. Now the arm knows something is wrong, and it needs to visit other locations to locate block 2 and move it to location C (which is an extra action which is not included in the original plan). If this time nothing is wrong, the arm can proceed to move block 1. If block 2 drops again, then the arm needs to determine another extra action.

Note that the light sensor provided in the LEGO Mindstorms Kit can tell the difference among three light intensities: white, black and green. So to be recognized by the light sensor(s), the objects need to have one of these colors.

If your robotic arm is able to perform online planning, your team will be awarded 20% bonus marks, which will make up for whatever marks deducted from other aspects of this homework. The maximum final mark you can get is still 100%.

Resources

1.     LEJOS: http://lejos.sourceforge.net/

2.     LEGO Mindstorms Homepage: http://www.legomindstorms.com/

3.     LEGO Robotics resources: http://www.crynwr.com/lego-robotics/

4.     Local Lejos and Mindstorms set-up instructions from Weihua

Min-Yen Kan <kanmy@comp.nus.edu.sg> Created on: Tue Feb 10 13:54:48 2004 | Version: 1.0 | Last modified: Mon Mar 22 09:46:08 2004

Pictures from their work

Group

Robots ...

... with their creators*

Group 35:
Karl Dailey
Danesh Hasan Kamal
Lau Hon Sun

Group 18:
Cheng Youjiang
Lau Qiangfeng,Peter
Lee Shugang

Group 25:
Naganivetha Thiyagarajah
Shakith Devinda Fernando
Sumedh Chalermkanjana

Group 24:
Ang Wee Cheng,Marcus
Huang Chi-Kai
Liu Siliang

 

Final Exam Questions

I’ve included two questions from the final exam.  The first shows that the students have to think about the principles and not just remember the facts taught to them.  The second shows that even calculation questions can be made interesting. 

 

Also, I’ve included a comment from two students about the final that I think are particularly helpful.

2.               Genetic Algorithms (8 points)

 

The genetic algorithm has two steps in generating an individual from a population.  The first step selects two individuals at random with probability proportional to their fitness.   The two individual parents reproduce by contributing parts of their genome to create a new individual.  In a second step, the new individual’s genome is possibly mutated at some random location in its genome.

 

 

Briefly discuss the impact on the algorithm if the second step of the (mutation) step is omitted.

 

8.         Learning from Observations (20 points)

 

Part I. Ms Apple I. Podd is an SoC student who listens to music almost everywhere she goes.  She often has homework due for her modules and many of them involve programming.  We’ve sampled her choice of music genre over several points in time. 

 

Using information gain, compute the decision tree for the following observations of Apple.  Show all work.  The following mathematical formulas may be useful for your computation. 

 

[N.B. Data rows omitted for space]

 

From:

Ty (nickname)

Date:

12/02/2004 08:25:00 PM

Topic:

Re: Exam format Query

 

 

Just making sure..

Anyway, judging from the midterm questions, I guess this module requires more thinking skills than memorising skills. So I guess its alright even with just an A4 sheet.

From:

LEONG WAI KAY

Topic:

Re: last post

 

Yah he prolly had an AI agent mark those papers.. lolAnyone wanna discuss some of the questions from the exam? Think the questions were rather intresting and thoughtfully set.


Summary

In this module, I accomplished my first semester of teaching a large undergraduate class by myself.  I hope to improve the curriculum in the future by tying together the (interesting) homework assignments I have created more closely with the course curriculum.

 

The first avenue for improvement is to re-do the tutorials as they have been used in the course for many times.  The students know old material when they see it and the questions are not relevant as they need to be.

 

From:

TEE WEI IN

Date:

25/04/2004 11:42:00 PM

 Heading:

Default Heading

Topic:

OMG!

 

Wow! SoC ppl truly truly amaze me.... This has been the best SoC module I have taken in my 4 years. Thanks Min, Weihua, and all those geniuses out there! You make me want to join SoC... if I could...

Man... AI meets human intelligence... a very lethal combination indeed....

 

Kathy McKeown 

to kanmy

 More options

 Aug 31 (4 days ago)

Min -- can you send me information about your AI class?

I'd love to be able to see:

   1. your syllabus
   2. your notes for class
   3. your assignments
   4. your tests

To start, I'm more interested in 1 and 3. Particularly 3, as I'm trying
to decide what game playing program to give them and I remember that
you had an interesting one.

Thanks!
Kathy

 

Samples from CS 6210 / 5244 – Digital Libraries

In Digital Libraries, I teach a course that is more oriented to the application of databases, HCI and text processing.  It integrates all of these fields yet needs to be balanced such that it does not significantly overlap with the materials of these other courses, taught by other professors.

 

The course is project-based and has a heavy emphasis on building core research skills is emphasized.  As such, a survey paper, presentation and final project are the core components.  Students are encouraged to make their work conference publishable.  At least one of the ten teams in the course did end up submitting their work for publication.

Course Description

This module is targeted to advanced undergraduates and beginning graduate students who wish to understand real-world issues in building, using and maintaining large volumes of information in digital libraries.

Fundamentals of modern information retrieval will first be taught, with a particular focus on how this fundamental technology is merged with traditional information finding skills of the librarian / cataloger / archivist.

Students will round out their knowledge with case studies of how different disciplines (e.g. music, arts, medicine and law) impose different search, usability and maintenance requirements on the digital library.

More detail on this course. Why have this course? Who should think about taking this course? What are the prerequisites? What are the aims and objectives of the course? How are students to be assessed?

Aims and Objectives

Modes of Learning

(Updated Thu Jul 24 14:15:05 GMT-8 2003)

Lectures and class participation will be used for the first half of the course. The first half will cover fundamentals of information retrieval and digital library standards, as well as furnish the students with a wide variety of topics of current interests to information professions in academic and industry.

Survey paper. By the third week of the course, students will have selected one topic (from a controlled list) to write up a survey paper on, using papers suggested by the lecturer and through their own information gathering.

Mini lecture presentation. Based on the topic of your survey paper, you will be grouped into small teams to present the material on the papers in your survey. These lectures will be in the last three sessions of our regular class schedule. Lectures will be 15 minutes long (about 5-8 slides), with five minutes for questions.

Project. After the survey, students will work on a particular project topic for further study, either individually or in groups of two or three. Students can either propose projects to me or choose from a list of topics. Our last class will feature a poster presentation (open to the public) which will allow the SoC staff and students to see the efforts of your work. Information on choosing a topic for the final project is given here.

N. B. - I know students are sometimes at a loss on how to assess their progress in a course due to lags in instructors or tutors returning assessment to them. I will try my best to grade work within two weeks of receiving the homework assignments.

Textbook and Required Readings

Required:

Recommended:

(Updated Thu Jul 24 14:07:58 GMT-8 2003) I have put a number of books that will be relevant to your project work in the RBR.
Click here to see the current books on reserve and their status.

We will also be reading selected papers from a number of relevant conference papers to give you an idea of the current state-of-the-art. You will be reading a number of additional papers as part of the survey paper.


Syllabus

(Updated on Tue Oct 7 14:50:20 GMT-8 2003

Below is the syllabus for the course. Not detailed enough for you? Try the complete, detailed schedule. N.B. - The notes below are the public copies, which are purposely incomplete. Students in class should use the versions in IVLE, which are more complete.

[ Web Information Seeking (Presented by Yang Hui and Guo Shuqiao) ]
[ Peer to Peer Systems in the DL (Presented by Li Yingguang and Vorobiev Artem) ]
[ Document Clustering for DLs (Presented by Lin Li and Wong Swee Song) ]
[ Intelligent Agents in the DL (Presented by Tok Wee Hyong and Qiu Long) ]

[ Spatial and Temporal (Maslennikov Mstislav and Chan Yee Seng) ]
[ Music and Speech in DL (Wang Gang and Huang Wendong) ]
[ Multimedia Mining and summarization (Edward Wiyaja and Hendra Setiawan) ]
[ Metadata Extraction (Chen Xi and Wang Xiaohang) ]

Reading Period: 15. 5 Nov 2003

Teaching-Learning Materials

I’ve illustrated some of the techniques I’ve used using only one lecture as an example (HCI in digital libraries).

 

Here I’ve used some strategically placed blanks to keep student’s attention and to force them to remember what they’ve heard.

 

 

I give real-life examples and annotate them on the slides so that students can follow both in the lecture and their notes (they get the .ppt or .pdf files of the slides) when studying or reflecting on the course.
 
I frame my lectures with introductory and conclusion material that is summarizes the information that I’ve given in lecture.  The questions that I pose at the end of lecture I also post to IVLE to get students to bounce the problem back and forth between each other.  The third bullet on the right hand slide generates my post in IVLE that I would like students to reflect on.
 

 

 

This module also included some presentations by students as part of their continual assessment.  I’ve included some of their work here  to illustrate the mode of peer to peer teaching.  They were given 20 minutes of presentation and budget about 10 slides per team.

  

 

 

 

Continual and Final Assessment

This module did not have either a midterm or a final.  As suggested by the Head of the Department, I have revised this module to have both a midterm and final, eschewing the continual assessment.  I have included one of the two assignments here.

Homework 2 - Authorship Detection

Quick Links: [ Home ] [ IVLE ] [ Project Info ] [ Schedule ] [ Details ] [ HW 1 ] [ HW 2 ]

As per our lecture materials on authorship detection, you will be creating machine learning features to be used with a standard machine learner to try to guess the identity of the author. We will be using papers from reviews of books from Amazon.com to try to compute attribution.

Details

You are to come up with as many features as you can think of for computing the authorship of the files. "Classic" text categorization usually work to compute the subject category of the work. Here, because many of the book reviewers examine similar subjects and write reviews for a wide range of books, standard techniques will not fare as well. You should use features that you come up with, as well any additional features you can think of. You can code any feature you like, but you will be limited to 30 real-valued features in total. You do not have to use all 30 features if you do not wish to.

In the workbin, you will find a training file containing reviews of books and other materials from 2 of Amazon's top customer reviewers (for their .com website, Amazon keeps different reviews in different counties -- e.g., UK). It is organized as 1 review per line, and thus are very long lines. Each line gives the review followed by a tab (\t) and then the author ID (either +1 or -1). There are 100 examples per author in the training section.

In the test.txt file you will find a list of reviews, again, one per line, but without the author ID given.

We are going to use the SVM light package authored by Thorsten Joachims as the machine learning framework. You should familiarize yourself with how to apply SVM to your dataset. SVM has a very simple format for vectors used to train and test its hypothesis. I should have demonstrated this during class. Be aware that training an SVM can be quite time-intensive, even on a small training set.

What to turn in

You will upload an X.zip (where X is your matric ID) file by the due date, consisting of the following four files:

  1. A summary file in plain text, giving your matric number (as the only form of ID), containing the percentage precision and recall for each class, under leave-one-out validation. You should inventory the features used in your assignment submission and briefly explain (1 sentence to 1 paragraph) explain the feature, if non-obvious. (filename: X.sum, where X is your matric ID). You can also include any notes about the development of your submission if you'd like.
  2. Your training file (X.train), consisting of the features for each review, followed by the author ID. This should be in a format that can be passed directly to the machine learner for training. Remember, you may only use up to a maximum of 30 features.
  3. Your testing file (X.test), consisting of the same features as above, but without the class labels (as you don't know them).
  4. A model file (X.model) that was induced by running svm_train with defaults (i.e., no arguments) to the machine learner.

Updated on Wed Oct 15 15:25:12 GMT-8 2003. Please use a ZIP (not RAR or TAR) utility to construct your submission. Do not include a directory in the submission to extract to (e.g., unzipping X.zip should give files X.sum, not X/X.sum or submission/X.sum). Please use all capital letters when writing your matric number (matric numbers should start with HT or HD for all students in this class). Your cooperation with the submission format will allow me to grade the assignment in a timely manner.

Grading scheme

Your grade will be 75% determined by performance as judged by accuracy, and 25% determined by the summary file you turn in and the smoothness of the evaluation of your submission. For example, if your files cause problems with the machine learner (incorrect format, etc.) this will result in penalties within in that 25%.

Of the remaining 75%, 45% of the grade will be determined by your training set performance, and 30% determined by the testing performance. For both halves, the grade given will be largely determined by how your system performs against your peer systems. I will also have a reference, baseline system to compare your systems against, this will constitute the remaining portion of the grade.

Please note that it is quite trivial to get a working assignment that will get you most of the grade. Even the single feature classifier using a feature of review-length-in-words, would be awarded at least 40 marks (the 25% plus 15% worse than baseline performance). I recommend that you make an effort to complete a baseline working system within the first week of the assignment.


Summary

In this module, I accomplished my first semester of teaching a graduate class.  This semester, I am re-running this course and am working on improving the curriculum by introducing more lectures on related topics (e.g., new media, blogging, instant messaging and short message service) to make the course more relevant to the students today.


Selected Student Feedback

Although you do have a full inventory of the student feedback on each faculty member at your disposal, I have provided a distilled version that highlights both weaknesses and strengths of my teaching.  I hope that these are valuable to you.

CS 1102 Data Structures (Semester II 2002/2003)

In my first semester of teaching, I co-taught 1102 with Sun-Teck, Wei-Tsang and David.  Sun-Teck and I were paired as a team to teach one section.  I taught the first half of the course.

 

Please note, this was the first year we fielded the online judge for grading of labs.  As such, there were many snafus (e.g., downtimes and crashes of the judge). This, in my opinion is the cause for the low student feedback ratings.

 

Overall instructor score: 3.715

Number of students: 205

Nominations for teaching award: 15 (7% of class)

 

Comments (positive, then negative)

 

CS 6210 Trial Run: Digital Libraries (Semester I 2003/2004)

This was my first run of my graduate course on digital libraries, coded as Special Topics in Computer Science.  It was small class, and as such, was easier to teach and also had less feedback.

 

Overall instructor score: 3.715 / 5

Number of students: 15

Perceived Difficulty (new for 2003/2004 reports): 3.4 / 5

Nominations for teaching award: 5 (33% of class)

 

Comments (positive, then negative)

 

CS 3243 Artificial Intelligence (Semester II 2003/2004)

This was my first attempt at teaching AI.  As mentioned already, we did a large revamping of the curriculum.  Working with Weihua, my TA for the course, we changed all of the homework assignments, as you can see in the module folder.  Students pointed out that the homework was not well connected to the syllabus and that the tutorials were too old.  We plan to re-examine these weaknesses in the next run of the course.

 

Overall instructor score: 4 / 5

Number of students: 154 (not counting PG students taking for QE)

Perceived Difficulty (new for 2003/2004 reports): 3.8 / 5

Nominations for teaching award: 5 (33% of class)

 

Comments (holistic, positive, then negative)

 

 

Appendix A – Emails on slide development

Before my conversion of the slides to PowerPoint format, there were many instructors on the AIMA mailing list looking for this exact resource.  I decided to do something about it.  As Stuart Russell (the textbook author) mentions in his email, no one had done this for the community yet.

203

Powerpoint slides of (edited) AIMA available

Min-Yen Kan

 knmnyn

Mon  1/5/2004

191

Re: Digest Number 85

Gregory Harrison

ploutous

Tue  10/7/2003

190

Re: how to use figures in powerpoint

Thomas G. Dietterich

uueeo

Mon  10/6/2003

189

Re: how to use figures in powerpoint

Michael T. Cox

 

Mon  10/6/2003

188

Re: how to use figures in powerpoint

Dave Swayne

 

Mon  10/6/2003

187

Re: how to use figures in powerpoint

Stuart Russell

 

Mon  10/6/2003

186

Re: how to use figures in powerpoint

Dave Swayne

 

Mon  10/6/2003

185

how to use figures in powerpoint

x2zhang

mulanzhang

Mon  10/6/2003

175

Preparing slides in Microsoft PowerPoint

m sad

sadsucc

Sun  9/7/2003

41

Re: PowerPoint Slides

Roy M. Turner

 

Fri  11/9/2001

40

Re: PowerPoint Slides

simi

 

Thu  11/8/2001

39

Re: PowerPoint Slides

GURKEWITZ@W...

 

Thu  11/8/2001

38

Re: PowerPoint Slides

Lisa Cingiser DiPippo

 

Thu  11/8/2001

37

Re: PowerPoint Slides

GURKEWITZ@W...

 

Thu  11/8/2001

36

PowerPoint Slides

Lisa Cingiser DiPippo

 

Thu  11/8/2001

26

PPT Version of AIMA slides?

Austin Texas

 

Fri  8/31/2001

 

Hi all:

> It would be nice if the publisher (or a volunteer) would produce a
> powerpoint version of the slides...

> Stuart Russell

First, thanks to the authors of the book for creating a wonderful
2nd ed. of the AIMA text. The supplementary material is very helpful.
For many of us, a powerpoint version of the slides would be helpful*.
I've translated (edited) chapter slides into powerpoint format
(chapters 1,2,3,4,5,6,7,8,9,13, 14 Section 1-2,18 Section 1-3).
You're welcome to take them and use them in your courses. I've posted
the web page versions and .pdf files on the web, the source powerpoint
files are available from me directly.

Please see http://www.comp.nus.edu.sg/~kanmy/courses/3243_2004/ for
the my syllabus and links to the powerpoint slides.

Details on slide translation:
- main TeX translated into tabbed text using a perl script
- imported into powerpoint and manually edited to add font changes,
symbols and equations
- algorithms' and posted as bitmaps from screen captures of the pdf
file. (I don't think many people will want to edit the algorithm
details...)
- figures and graphs transformed to gif using un*x convert image
conversion package. Figures and graphs then manually embedded and
aligned.
- no animations come with the default powerpoint slides.

Regards,

Min-Yen Kan
Assistant Professor
School of Computing, National University of Singapore

 

From: Juan J. Flores  <juanf@zeus.umich.mx>

 Dear Professor Min-Yen KAN,

Got the slides.  They work just dandy, except for some sqares that
appear at the end of some sentences.  Like the squares at the end of a
theorem proof in a book.

Anyway, they look like they will be of great help.  I intend to
translate them to Spanish.  It may not be of much help to you.  I will
let you know once they are done.

Best Regards,

Juan

 

Date: Tue, 6 Jan 2004 13:38:47 -0800 (PST)

From: Colleen van Lent <vanlent@cecs.csulb.edu>

To: kanmy@comp.nus.edu.sg

Subject: Powerpoint slides

 

        I am the instructor for AI at California State University Long

Beach.  I would appreciate the original zip file for your slides. Thank

so much for taking the initiative to finally make some powerpoint slides.

 

        Thank you,

 

                Colleen

 

Date: Wed, 7 Jan 2004 14:14:58 +0000

From: Irfan Awan <i.awan@scm.brad.ac.uk>

To: kanmy@comp.nus.edu.sg

Subject: Re: [aima-instructors] Powerpoint slides of (edited) AIMA available

 

Dear Kan

 

Many thanks for your email and providing me the powerpoint slides. Your

efforts are marvelous.

 

Kind regards

 

Irfan

Appendix B – IOI / ACM Training Slides

Here are some slides from the training session for our IOI / ACM teams.

 

In this slide, I relate the idea of comparison sort to three basic sorting algorithms.   I show that the three are basically the same algorithm with different specifics.

 

I like to intersperse humor into my slides to perk them up (both the slides and my students!)

 

I also taught invariants, a nice trick to show that what appears to be a search problem isn’t always a search problems.  Look for solutions beyond the brute force method.   

Appendix C – Teaching Practicum report excerpt

 

I’ve edited this report for conciseness and to highlight the teaching contributions of the practicum research.  “…” indicates locations where I have edited the report.

 

Constructing Homework Options

Dr Kan Min-Yen / Department of Computer Science

 

The freedom of choice can motivate students to perform better in their coursework, by letting them choose topics of interest or, equally, to avoid topics of disinterest.  However, the power of choice is often not a factor in assessing student performance: tests must be taken, homework must be turned in by a deadline.  This need not be the case, as the teaching staff can offer options for student assessment.  We present a case study along this line, in which we offered different options for a homework assignment. In our case, the incorporation of the different options led to a restructuring of the curriculum of the course, but resulted in positive impact both with students and instructors.  In presenting the study, we focus on answering four questions regarding homework options: 1) when does it make sense, 2) how do students choose which option to pick, 3) is it rewarding for students, and 4) is it rewarding for teaching staff?

 

We studied the effects of homework options in the context of the Foundations of Artificial Intelligence course offered at the School of Computing.  The course is a third-year elective course that presents basic material needed for advanced coursework in A.I.  Based on past enrollment, most students who take this module do enroll in advanced course work later on in their schooling.  With this in mind, we decided to have students chose between different homework options that would give them a taste of an advanced course. 

 

Three different options for the assignment were drawn up, each giving a simple problem within the advanced topics: robotics, computer vision and natural language processing. While this expanded the curriculum beyond the original scope of the course, it was hoped that the assignment would provide a concrete grounding of the foundation materials given in the course, encourage students in enrolling in an advanced area, and motivate students to work on a topic of their interest. The assignment would allow them to investigate an area without investing a whole semester and in a way that would not significantly overlap with the advanced course’s curriculum.

 

The homework options themselves required that students to master additional material on top of the standard lecture material.  Following the one hour midterm, we introduced the homework options and asked students to watch one of three supplementary lectures that we pre-recorded and offered by webcast.  The assignments were done in groups of two or three, which students formed on their own.

 

Successful scenarios for homework options

 

The success in offering homework options depends on a number of parameters.  In preparing our options, balancing the difficulty between homework assignments seemed to be the most important factor.  Other factors included whether a complete specification of the assignment was necessary and whether detailed grading criteria was given.  In addition, we also analyzed whether a strong connection with the foundation course material was necessary.   … Among the factors that we surveyed, students agreed (4.2 of 5) that homework options need to be interesting.  A comment from a student puts this nicely: “Since I chose this assignment out of interest I feel that I was able to perform better in it (i.e., it was more fun to devote time on it) than something forced on me)”. 

 

A second observation is that students slightly prefer assignments that are directly tied to the course syllabus (3.4 of 5).  This was a bit disappointing to us, as our homework options did not directly correlate with the standard lecture materials (which were to be assessed in the final exam).  Students also slightly agreed (3.5 of 5) that assignments should have completely specified grading criteria but agreed less (3.2 of 5) that the assignments themselves need to be completely specified.  Students’ comments on this aspect were quite constructive: “[the] lack of specific makes the judgement [sic] of effort difficult…”, “the assignment problems should be publish [sic] earlier so that we can prepare some reading in advance”, “I like the idea that one part of the assignment is left open for us with no ‘model answer’. This allows us to express more creativity. However, due to heavy workload, a more open-ended assignment tends to be more heavy”.   These comments show the correlation between open-ended assignments and workload.

 

How do students choose which option to pick?

 

One hypothesis that we wanted to disprove is whether students would pick the option that seemed least difficult.  To counteract this, we attempted to balance the difficulty of the options, even though the end result was still slightly imbalanced in our view.  Fortunately, it seemed that no one assignment was particularly favored: the class was almost evenly split between the three options.  Post-assignment, students were neutral on this issue: perceiving the assignments to be neither balanced nor imbalanced in terms of difficulty (3.0 of 5).

 

…

 

Rewarding to students?

 

Students agreed that valued having the power to choose (4.1 of 5) which assignment to take.  They preferred having the choice of the current assignments (which by the students’ own assessments were difficult and open-ended) to having a single mandatory, easier and completely-specified assignment (3.3 of 5).  It is encouraging to see that among the students who participated in the survey, many prefer the challenge and customization of the course to their interests to an easy and comfortable alternative.  Students comments reinforce this conclusion: “Having the liberty to choose is definitely a positive aspect as it enables us to align our interets [sic] with the course syllabus”, “Also, I liked that I could choose between topics, and that the assignments took the learning outside of the lecture material. I think that is what a rounded education should be.”, “I thought giving us the freedom to choose one topic from a variety is a very good idea. Because each student has different strengths: one may be stronger in [one area] while the other in [another area].”

 

Rewarding to instructors?

 

Any positive impact homework options have on student interest and drive to excel certainly contributes to a rewarding experience for instructors.  However, preparing homework options takes a toll on the teaching staff.  Is all the effort worthwhile? 

 

We tabulated the number of hours used in preparing the homework options for the course, breaking hours into mandatory time as well as optional time that we invested beyond what was absolutely necessary for the options.  A total of 46 hours was spent in the execution of the three options for our course.  In comparison, a single assignment may have taken less than half of the time.  To properly fit options into our course required structural changes: adding the webcast lectures and re-working the syllabus.  This additional overhead may be avoided in a basic setting, whittling the time cost to about 36 hours. 

 

Conclusion

 

Homework options definitely add value to a course.  Options allow individual students to attempt assignments that are of interest to them and this in turn encourages them to excel and invest energy to do well in a course.  Of the results of the study, the two strongest observations resonate with each other: 1) students value the power of choice in deciding which homework assignment to choose, and 2) homework assignments should pique the interest of the student.  These two factors can have a natural synergy too, as demonstrated by a student: “All three topics are interesting. It would be a much enriching experience if we have the luxury of time to attempt all 3 topics.”

 

Having the experience of offering homework options, we have some recommendations for those interested in using this pedagogical technique:

Appendix D – Beginning research presentation excerpt

----- Original Message -----
From: Steven Halim <g0403330@nus.edu.sg>
Date: Wed, 18 Aug 2004 14:45:08 +0800

Subject: Survey paper preference

To: Dr Min,
I've decided to do survey on the following topics, with Bioinformatics as the first choice (I'm taking CS5238 - Bioinformatics too anyway)...

1. Bioinformation and Genomic Data in DLs
2. New Media for DL - Blogging | IM | Wiki

Btw, through this e-mail, I want to ask permission for using adapted version of your slides in Lecture 1 regarding how to read scientific paper effectively. I'm going to include this in my Computational Logistics Lab seminar this Friday where I need to address some HYP students regarding how to approach their HYP-year effectively. Is this okay?

Regards,
Steven Halim

I go over the principles to success for HYP and UROP projects with all of my beginning undergraduate project students.  Several others have borrowed and adapted my slides.  I include an email from one below [emphasis added].

                                               

 

I go over what happens in a HYP / UROP presentation so that they can get a feel for what’s to come…


 

 

And give them tips about how to do research.  These things are obvious to us now, but to a beginning student they are often very much appreciated.

 

Appendix E – Graduate Research Course Joint Poster Seminar

I have run a graduate research course project poster session for two different semesters.  The purpose of this poster session was to organize all of the research project based graduate course work happening during a semester into a single presentation session.  This would foster graduate student camaraderie, and encourage students and staff to learn about each other’s projects and interesting course work.  Each graduate course would mandate that their students prepare a short poster presentation on their research topic.

 

This was an initiative started by me, and has not be done by any other staff members in the department.  I orchestrated the groundwork early in the semester to ensure that graduate course instructors using projects could participate in the seminar, and worked alongside them to plan the logistics, consisting of the venue, food and getting departmental support.

 

The first joint poster session featured five different lecturers in the computer science department, held in the Fall of 2005 (Sem I).  The second session was held in 2006 (Sem I), featuring four lecturer’s courses and a best poster awards session.  I have archived one of the poster sessions online at the below URL, which I have given a partial screenshort below.

 

http://www.comp.nus.edu.sg/~kanmy/courses/poster_session_sem1_2006/

 

Appendix F – Examples of Discussion Questions

 

As stated in the Main Dossier  (see Teaching Philosophy A.1), I use both lecture and IVLE to introduce thought questions that frame the lecture materials in the larger, real-life context.  After each class, I re-post these questions on IVLE.   As these are open-ended questions, the students’ answers help themselves to sharpen their viewpoints, to help teach and inform their peers, and help me assess their understanding and interests.

 

Here are two examples from this semester (Sem 1, 2008/09 – Text Processing on the Web) of closing discussions post that I had with the students.  You can see this is quite a lively forum despite only having about 30 students.  You can also see that there are quite a few responses to these posts, mostly from students contributing to the forum.  I had many more such examples but due to IVLE migrations, I’ve lost the other examples.