ME290M Expert Systems in Mechanical Engineering Spring 1999, T-Th 12:30-2:00 pm 1165 Etcheverry Hall, Course Control No. 56369 http://best.me.berkeley.edu/~aagogino/me290m/s99

	Frequently Asked Questions on Assignment #4. (1) On problem #2 (4.12 in the text): You are to assume that E1 and E2 are conditionally independent given H. Thus Pr(E1|H)Pr(E2|H) = Pr(E1,E2|H).
	(2) On problem #6: The answer should be a function of the probability of winning "p". If p>0.5, would you play the game if you are an expected value decision maker? In fact you would invest all of your money. If you had the logarithmic utility function given then the Utility for winning would be ln(W+X) and the Utility of losing would be ln(W-X) and the expected value of the Utility would be p*ln(W+X) + (1-p)*ln(W-X). If you want to maximize the expected utility you would need to optimize for X (differentiate with respect to X -- back to finding minima in calculus). The extra credit has to do with how you might change your optimal investment given you could play the game twice, thrice or even an infinite number of times. (3) A number of students in office hours today had a question on Problem #5 (or 5.11 in the text). The answer is entirely subjective. You should define your own linguistic variables for part (a) and then draw functions for (b). I don't see how you can do it for part (c) without being subjective as well. So I'll accept any reasonable answer here. I just want you to go through the exercise of defining your own linguistic variables for uncertainty. (4) On the presentations: You should make this informal. I would prefer not to do computer projection as it takes so much time to set up and it will be hard to move between students. You could just stand up and summarize without any visual aids. Or you could use an overhead if you want. But let's keep this informal.

We will have two guest speakers on Tues., May 4. First will be David Berman from Knowledge Industries. He will talk about on-line diagnostics in the automotive industry and a working demo of several Bayes networks using their web hosted inference engine and a downloadable software set. The second speark will be Prof. Marti Hearst in the School of Imformation Management and Systems (SIMS) speaking on her work in text data mining. She is also the developer of the search engine "cha-cha" used on the the Berkeley campus home page and recently features in an article in the Journal Science ("Assembling the World's Biggest Library on Your Desktop", 1998 September 18; 281: 1784-1786). Also see Prof. Hearst's paper "Distinguishing between Web Datamining and Information Access. .

As Laura pointed out, problem #6 on assignment #4 was duplicated from assignment #3. I wouldn't want to short change you on homework, so I have added a new problem that is quite interesting. The new problem #6 is a variation of what is called the St. Petersburg Paradox. You only need to do part (a). As a few of you have asked for some extra credit to pull up your grades (those overconfident students who put 0 probability on the right answer, for example), I have added a parts (b) and (c) to this problem for those folks or anyone interested. The modified assignment #4 is at:

	Our guest speaker on Tues., April 27 will be Prof. Theodore E. Cohn, Professor of Vision Science, speaking on "Human Vizualization". Prof. Cohn suggested students contact him directly to get a tour of his human vizualization lab. His office is 360 Minor Hall, UC Berkeley, Berkeley, CA 94720-2020; Phone: 510-642-5076; Fax: 510-643-5109; Web: http://vision.berkeley.edu/VSP/TEC/.
	Our guest speaker on Thurs., April 22 will be Andreas Friis-Hansen from the Department of Naval Architecture and Offshore Engineering Technical University of Denmark and a student in our class. The title is: Inspection Planning using Bayesian Networks. He will also solve the "Oil Wildcatter's Problem" in class using the Hugin Bayes' Net software.

	Prof. Lotfi Zadeh will be our guest speaker on Tuesday, April 13, 1999. He will speak on fuzzy logic and soft computing. In addition to the text book reading, a good reference is his IEEE paper titled "From Computing with Numbers to Computing with Words -- From Manipulation of Measurements to Manipulation of Perceptions," Lotfi A. Zadeh, IEEE Transactions on Circuits and Systems I - Fundamental Theory and Applications, Vol. 45, No. 1, Jan. 1999, pp. 105-119. (Search over database).

	A number of you have asked questions concerning problem set #3. It appears that the wording on some of the textbook problems are a bit ambiguous. I thought I would respond to the entire class on these Frequently Asked Questions (FAQs). I have also used this clarified wording on the version of the homework assignment #3 on the class website.
	Problem 4.3 from text (My problem 3). You can assume that the probability of good or bad components in each bin is the percentage of each in each bin. Problem 4.4 from text (My problem 4). You are given that the probability of a crash IF Brand X is chosen is 0.1. And so on for each Brand. For Part (b), you are asked the probability that Fred's drive crashes in a year. This would be the joint probability of Fred choosing a brand and the event that this drive crashes. On part (c) you are to give the probability that any drive tht Fred owns crashes in a year. Thus the probability of a crash in any one year is not 100%. It is the sum of the joint probabilities of the events that any drive is chosen and its failure rate. Problem 4.9 from text (my problem 8). I believe the authors wanted you to find: Pr(F1|A), Pr(F1|B) Pr(F2|A) and Pr(F2|B). Anyway that is what I want you to give me. I'll give you 5 points extra credit if you can also give me: Pr(F1|C) and Pr(F2|C), where C= event that of {bad writes}.
	How to send Jiangxin your CLIPS programs. save your original programs ( define rules, templates, etc ) by using a filename that start swith the initial of your first name, followed by your last name, then followed by 2_5p.clp for prob(5) and 2_6p.clp for prob(6). (e.g., jwang2_5p.clp and jwang2_6p.clp for me) save all your constructs after you run your rules under clips to files with names that start with initial of your first name, your last name, and then 2_5r.clp and 2_6r.clp. ( use 'save' command to do this under clips ) (e.g., jwang2_5r.clp and jwang2_6r.clp for me) If you are having trouble saving your file,
	Here is a copy of the CLIPS exercise presented by Jiangxin Wang in class on Thurs., Feb. 25. It is problem 7.10 from the text.
	Card keys for the Mac Lab are available from Joey Cruz in 2115 Etcheverry Hall. If you already have a card key for Etcheverry Hall he just needs to validate it for the Mac Lab. There should be no "use fee" for this service, but there may be a small service charge to pay for the processing.
	The X-Windows version of CLIPS is now on newton. Just type 'xclips' for the X-Windows version and 'clips' for the regular version to run it. Please let Jiangxin know if you have problems.
	The due date for Problem Set # 2 has been changed to Thurs., March 4, 1999. One hint on #1. The representation that you use can make a big difference on this problem. I suggest you "follow the hole" and the moves around the hole rather than developing a more complex representation.

Thurs., Feb. 25, 12:30-1:00pm our Expert Systems class will meet in 3117b, the Sheppard Room, in order to talk with Dr. Kent Cullers from SETI (Search for Extraterrestrial Intelligence) about their approach to identifying intelligence in space.

The following bio was taken from a SETI brochure: Blind since birth, Kent Cullers has devoted his career to seeing beyond Earth. The first totally blind Ph.D.-holding physicist in the Unites States. He has played a pivotal role in SETI, devising ingenious schemes for ferreting out faint signals from a large and growing sea of noise. After earning his bachelor's degree in physics from Pomona College and his doctorate in physics from UC Berkeley, Dr. Cullers began working in support of NASA's SETI program and then the SETI institute after Congress cut off funding for SETI with NASA. The recipient of many honors and awards, including the 1994 Federal Employee of the Year award, and the NASA Exceptional Engineering Achievement medal in March 1993, Dr. Cullers is a member of the American Astronomical Society, a board member of the San Francisco Lighthouse for the Blind and Visually Impaired and Chair of AAAS Committee on Opportunities in Science.

Dr. Cullers was the inspiration for the character Kent Clark in the film Contact. The SETI site maintains "the movie Contact" web site that describes the movie and compares the movie with reality.

	In class on Thurs., Steve asked if "none of the above" is an acceptable answer for question 5(a) on problem set 1. As I said in class there does seem to be a typo on this question. So yes, "none of the above" is a fine answer, just give an explanation. Or you could change the question to: There is a women that is alive and mortal.
	CONCEIVING ADA SITE. In my history of AI lecture, I said that Ada Byron Lovelace is often credited with writing the first computer program and being the first AI philosopher. It turns out there is a new movie about her called CONCEIVING ADA, which B. Ruby Rich listed as ONE OF THE TEN BEST FILMS OF THE YEAR. It will be first released in San Francisco if anyone is interested on February 19 at the Lumiere in San Francisco. I can't comment on the historical accuracy or the quality of the film, but have attached some PR about it below for those who might be interested.
	"Riveting" Sid Aidelman, Toronto Star The film won OUTSTANDING ACHIEVEMENT IN DRAMA at the Electronic Cinema Festival in Japan, and has screened at approximately 35 festivals worldwide, including The Toronto International Film Festival, The Sundance Film Festival, Riocinne, and The Berlin Film Festival (where it was the image for the festival poster). CONCEIVING ADA is an imagined biography of Ada Byron King, Countess of Lovelace, played by TILDA SWINTON. The daughter of the poet Lord Byron, the mathematics genius, developed in 1832 what became the world's first computer language. The film features KAREN BLACK, TIMOTHY LEARY (in his last performance), JOHN PERRY BARLOW and original music by THE RESIDENTS. The film is produced by Henry S. Rosenthal. This is the first film to ever be produced with virtual sets, a process that was invented and patented for the film. THE OFFICIAL CONCEIVING ADA SITE http://www.ted.net/ada/ CONCEIVING ADA'S LISTING AT THE SUNDANCE FILM FESTIVAL '98 http://www.sundancechannel.com/festival98/filmguide/films/conceivi.tin.
	Tuesday, February 9, 4:00-5:30, Andersen Auditorium. "New Strategies of Venture Capital Investing and Their Impact on E-Commerce," Geoff Yang, General Partner, Institutional Venture Partners.
	Recent developments in the stock market have changed the way venture capitalists think about the growth of high tech companies. With the huge returns realized on technology stocks in the past few years, venture capitalists are thinking big and searching for "homeruns". Valuations are skyrocketing and VCs are putting enormous sums of money to work in an attempt to gain first mover advantage and become the "gorilla" in newly developing industries. Geoff will discuss the impact of these changes on one industry: E-Commerce. Mr. Yang joined IVP in 1987, having previously been an Associate with First Century Partners. Earlier, Mr. Yang held marketing and sales roles at IBM Corp. Since joining IVP, Mr. Yang has emphasized communications investments in data networking and Internet infrastructure. He currently serves on the Boards of Atmosphere, Excite (XCIT), GWcom, MMC Networks (MMCN), Promatory Communications, Shasta Networks, SkyStream, TiVo, Turnstone, WaveSpan and Whistle Communications. He has also led IVP's investments in AlacriTech, Foundry Networks, Juniper Networks and Netcore. Previously, he had served as Director or was actively involved with Applied Digital Access (ADAX), Centillion, Collabra, Crescendo, Golfweb, Netlink, Rapid City, Synernetics, SynOptics (SNPX), Teleos, Wellfleet Communications (WFLT) and Whitetree. In addition, Mr. Yang serves as a Director and immediate past President of the Western Association of Venture Capitalists, Director of the National Venture Capital Association and Chairman to the Stanford Engineering Venture Fund. Mr. Yang holds a B.S.E. in Engineering Management Systems and a B.A. in Economics from Princeton University and an M.B.A. from Stanford University.