ECE 579 Intelligent System

Fall 2001

Course information:

Syllabus                    Term Projects

Project training data:

Data Set 1                 Data Set 2

95 Data Set 1            95 Data Set 2

Project testing data:

95 Data Set 3             Test Data Set

 

Course Syllabus

Professor

Textbook

Artificial Intelligence, George F. Luger. Addison-Wesley, 4^th Edition, 2001

Reference Books

1. Computer Systems That Learn, Sholom M. Weiss and Casimir A. Kulikowski, Morgan Kaufmann, 1991.
2. Machine Learning, Tom M. Mitchell, The McGraw-Hill Companies, Inc., 1997.

Topics to be covered

1. Introduction to intelligent systems in industry and manufacturing (1.5hour)
2. Machine learning(1.5 hour)
3. Basic search algorithms(4.5 hours)
4. Rule-based systems (4.5 hours)
5. Fuzzy logic(4.5 hours)
6. Neural networks(4.5 hours)
7. Decision tree(4.5 hours)
8. Bayesian learning(3 hours)
9. Advanced concepts in building and evaluating intelligent systems, and Case study: vehicle fault diagnostics (4.5 hours)

Grading

Statement on Academic Integrity  (Approved August 9, 2000)

The University of Michigan - Dearborn values academic honesty and integrity. Each student has a responsibility to understand, accept, and comply with the university's standards of academic conduct as set forth by the Code of Academic Conduct, as well as policies established by the schools and colleges. Cheating, collusion, misconduct, fabrication, and plagiarism are considered serious offenses. Violations will not be tolerated and may result in penalties up to and including expulsion from the University.

Back to Top

Term Projects

1. You are required to write a program that can learn from a training data to classify two classes. You are provided with a data set called TRAIN.txt. Each line in TRAIN.txt represents a data example and has the following format:
   The first 5 data are features used for classification, the last data on each line indicates class label that has one of the two values 1 and –1.

   Your program should read the TRAIN.txt and build a knowledge base.  After the training your program should be able to read in a file consisting feature vectors only.  For each data example in the input file, your program should generate a class label as the classification result. On the date of demo, you will be provided with a data set TEST.txt and your program will be tested on this data set.

   You can implement either one type of intelligent systems discussed in the class for this project.

2. You are given a set of data in the following format:
   The given time series of length T samples is produced by a physical system. Each sample k of the time series consists of four inputs and one output (k=1,2,…,T). The first input x1(k) represents a binary synchronization pulse related to a natural periodicity in the system. The second and third inputs, x2(k) and x3(k), represent context. The fourth input x4(k) has a more direct relationship with the bipolar output y(k) (the sixth or rightmost column), but the relationship is complicated due to dynamics of the system. The objective is to obtain a model of the underlying physical system (data generator) that will best describe the input-output behavior of the system. To evaluate the accuracy of the model, another time series is provided with the output column missing.

   Please keep in mind that there is an additional (fifth) column in the data files. It indicates samples for which the predictions are to be counted. Only if the k-th value in the fifth column is equal to one do we evaluate accuracy of the prediction of the output made for the k-th sample.

   You are suggested to do the following for your term project:

   1. Analyze the data and find the effective input feature vector to an intelligent system.
   2. Design a neural network to achieve high accurate classification rate.
   3. Generalization analysis: you are requested to build a committee of small neural networks with the given software to solve this problem. Your results should be compared with a large neural network in terms of generalization.  
3. Choose a problem that is related to the intelligent systems theme and you have work/experience in dealing with the problem and you will have data available for your program development and test.

WHAT TO TURN IN

You need to turn in one hardcopy of your report and a diskette that contains:
(1) text file of your report
(2) your program in both executable and source

On the demo date, you must attend the entire class demo and present your project and demo. You need to turn in your project report and the program you implemented on the specified due date. No late submission is allowed unless a medical emergency occurs, in this case a medical doctor’s letter must be submitted. The program should be well documented and tested properly. If for any reason I cannot test your program properly, your grade will be penalized.

Your project report must consist of the following sections (number your sections)

1. Introduction: describe what problem you intend to solve and what type of intelligent system you want to implement, and give an overview of the technology related to your chosen project.
2. Describe your system design.  A block diagram should be included.
3. Describe your program performance
4. Conclusion: what have you learnt from this project, what possible future work you think you need to do in order to improve your program, what are the open issues you discover that are important in the further development of this system.
5. Reference: You must use the same format as IEEE publications. If you do not know the format, you may go to the library to pick up any IEEE periodic publications, e.g. IEEE Computer,  IEEE Pattern Analysis and Machine Intelligence, IEEE Computer Networks, etc.

  Back to Top

Last Updated: 11/15/01