This is an advanced course in artificial intelligence (AI), covering a wide range of AI methods, concepts and applications. It is effectively "AI part II", a sequel to the AI intro course (W4701). The first portion of the course will be oriented around well-structured homework assignments, and in the later portion you will focus on implementing a larger, open-ended project. Note that this generalizes from the bulletin's and CVN's course description to expand the focus beyond expert systems per se (although the concepts of expert systems will be covered, and an expert system is an option for your final project).

Professor: Eric V. Siegel ( email: evs at cs dot columbia dot edu ). Office hours: Tues 4:30 - 5:30, Wed 2:00 - 3:00. office: 703 CEPSR, 939-7112

Eric Siegel now provides predictive analytics services at Prediction Impact, Inc., where predictive modeling and data mining are applied to gain customer intelligence.

TA: Jeff Commando Sherwin (jcs52@cs.columbia.edu). Office hours: TBA office: TBA

Framework of AI Approaches for this course:

• Symbolic AI and knowledge representation, such as:
  • terminological knowledge
  • concept subsumption inference
  • ontological engineering, e.g., temporal representation and reasoning
  • planning
• Probabilistic AI, such as belief networks
• Machine learning, such as:
  • neural networks
  • induction over belief networks
  • computational learning theory
  • unsupervised learning
• Applications, such as:
  • computational linguistics
  • data mining
  • expert systems, such as medical diagnosis
  • games
• Advanced and misc. topics, such as:
  • hybrid systems
  • artificial life
  • AI buzzwords galore

Homework projects:

1. Temporal constraint inference; application thereof
2. Neural networks for image classification, e.g., face recognition
3. Short oral presentation of a published research paper
4. Final project (optionally can build on earlier homework)

The last two projects above facilitate learning about AI research, including:

• Understanding research papers
• Performing experimental or system-based research
• Writing a report of research results
• Orally presenting research results

Course Themes

• The semantics of linguistics and temporal reasoning
• Representation issues: modularity, "Chomsky status"
• Overfitting versus generalization
• Evaluating results
• Meta-intelligence

Text: Readings will included research papers handed out in class, and a few later chapters from Artificial Intelligence: A Modern Approach, by Russel and Norvig, Prentice Hall, 1995. ISBN # 0-13-103805-2.

Grading : 16% HW1, 16% HW2, 12% midterm, 8% paper presentation, 31% final project, 17% final exam.

Late homework: Late assignments will be penalized. Note that partial credit will be considered for all incomplete work.

Collaboration: Discussion of material covered in class is strongly encouraged. Some assignments will be designated as collaborative. Otherwise, the work you submit must be your own work.

Open Door Policy: We would like the course to run smoothly and enjoyably. Feel free to let us know what you find good and interesting about the course. Let us know sooner about the reverse. See us, leave us a note, or send us an e-mail.

email: evs at cs dot columbia dot edu

INFORMATION FROM PREVIOUS OFFERING OF THIS COURSE:

STUDENT FINAL PROJECTS:

• NEURAL-NET HAND SIGNAL RECOGNIZER
• Theme Identification in Alice in Wonderland
• A Framework for ALife-Motivated Learning Experiments
• Two Network Face Recognition Using Unsupervised Training
• GENETIC PROGRAMMING IN NEURAL NETWORKS
• Application of Two Layer Supervised Back Propagation Neural Networks to Edge Detection in Still Images
• The Generic Genetic Problem Solver
• NFL Predictor
• An On-line Othello Program Trained by Genetic Algorithm
• Bus Recognition and Prediction (video camera out their apartment window looking for city buses)
• Learning Cost-Sensitve Classification Rules for Network Intrusion Detection using RIPPER
• Natural language temporal reasoning
• Crossword Puzzles as Constraint Satisfaction Problems
• GENETIC PROGRAMMING IN NEURAL NETWORKS
• Neural Network with forgetting algorithm on the prediction of stock prices

FINAL ANNOUNCEMENTS:

• Please copy your writeups, etc. into Eric's AcIS directory ASAP. Thanks.
• Eric needs hardcopies of final project writeups ASAP under his office door.
• Paper presentation grades are available off Jeff's course page (link below). Grades are out of 10 possible points (despite the fact that it is 8% of the semester grade). Grading was based on organization, prepared slides, demonstrated understanding of paper, conveyed basic content, and conclusions/personal response.
• Final exam -- due Monday, May 10.
• FINAL PROJECT WRITEUP ASSIGNMENT
• FINAL PRESENTATION ASSIGNMENT
• Guidelines for the final presentation are coming VERY soon.
• After you have written your final project up in HTML, put it and your presentation slides in my directory. On AcIS, I made ~es66/public_html/finalproj/ writable -- make a subdir with the last name of a team member, and copy your stuff in there. Make the main file called index.html (not .index.html). ALSO, please do so for your temporal writeup (if it is text or HTML) in /temporal/ (writeup and implementation!) and your NN writeup, if it is in HTML, in /nn/ -- THANKS!
• DEADLINES: final presentations are during the final exam slot, May 13 1:10-4. Final exam is take-home, and is due May 10, under Eric's door, 703 CEPSR. Final project writeup and final presentation are not due until the presentation time slot. But you are strongly advised to complete all implementation and/or experimentation by May 3, after which you can focus on doing the final, doing the final project writeup, and preparing the final presentation.
• If you want to do a project this summer or fall email projects@cs or see www.cs.columbia.edu/~projects -- cc me for my info.
• FINAL PRESENTATIONS: May 13 THURS MUD 545 110 PM 400 PM

The final exam is take-home, given out last class, due may 10. Final proj due may 13, but highly recommended that all but writeup is done may 3

Course syllabus -- What happened in the past; what is planned for the future.

Jeff's (the TA's) homepage for this course.

"Learn This" Rap (on machine induction)

HW1: Temporal inference

HW2: Neural Networks

Class newsgroup (columbia.spring.cs4721-sec1) (Eric just added this - email him and Jeff if it doesn't work.)

Course notes that I made for some of the lectures. These lecture notes are outlines that guided the talk -- you will not get much more than a general outline from them unless you have seen the lecture. But they can be helpful reviewing resources, e.g., for the midterm. Also note that within each lecture, the order of the slides is often messed up.