Announcements

May 15 The A6 test script and solutions are here. 10 points for the quality of your fitting function. Roughly full credit if almost all the max errors are <=10 minutes. Points deducted after that. 10 points for HowISolvedIt togeteher with the WriteUp.pdf (if included). Points deducted if codes/plots insufficiently annotated or did not run without adjustment. The solutions posted are authored by a student and representative of full-credit work.

May 12 Instead of 2:30-4pm Friday office hours in my office, I will hold them in 4135 Upson at the same time and call it a "review session." I will go over the review questions and anything else. Solutions to the review question will be posted shortly thereafter.

May 6 Slides for Friday's lecture are here. Study week office hours are here. Review question for the final will be posted on Monday as will the time and place of a review session. A6 should be graded by the middle of next week.

1. SunRiseAndSet should "hardwire' your fitting function. By that I mean that all its parameters should be written into the code. SunRiseAndSet should NOT call lsqnonlin.

2. The parameter values used in SunRiseAndSet should be reveal when HowISolvedIt is run. So among other things, it should have some ``final'' call to lsqnonlin that resolves the parameters of your fitting function. It should also produce plots and/or tables that shed light how you derived SunRiseAndSet and how you determined relevant starting values in your use of lsqnonlin.

3. The "sum" of the information in HowIsolvedIt.m and WriteUp.pdf is what is important. HowISolved it must be a working, nicely commented function. Use WriteUp.pdf to communicate whatever information is awkward to express in the form of Matlab comments.

4. You are encouraged to use elementary trigonometry.

5. Appreciate that the given data is accurate plus-minus one minute, so that is the "star to aim for." But use common sense when reasoning about error. The difference between 365.25 vs 365.2425 is about a minute. The Earth is a wobbling spinning top that completes a wobble about every 26000 years. 1/26000th of a year is a few minutes. Etc. My advice is NOT to use predetermined values like 365. Let lsqnonlin figure out the 'best" length-of-year value.

April 18. Prelim 2 covers Assignments 4 and 5. Here are some review questions.

Mar 10. Solutions to A3 (without partial credit guide) are here. The assignment will be graded by this evening and the partial credit solution guide will also be posted at that time.

Mar 3. In-class prelim on Friday, Mar 11. The Wednesday March 9 lecture will be a review session. Review questions for the Prelim are here. Office hours next week are different.

Feb 16 (a) You are allowed to use " \ " to solve triangular systems anywhere in the assignment. (b) The input vectors f and g in problem 4 should be column vectors. Note that the computational part of that problem requires effective use of lu. The written part of the problem involves SVD.

Feb 14. In problem three of assignment 2, a and b are not specified. Your function should assume that

Feb 10 Solutions to the first assignment are here. Send an email or come by if you think things were misgraded or if you have a question.

Feb 2. Practice uploading a .m file and uploading a .pdf file to cms. And make sure you know how to produce a write-up with a plot produced by Matlab. The due-dates are HARD deadlines. There will normally a short grace period measured in minutes, but do not count on it.

Feb 2. Regarding Problem 4 in the current assignment, here is a more accurate description of the IEEE double format than what is given in Chapter2.pdf. Use the facts in this write-up to do the problem 4. And just to be clear, count all the floating point numbers except NaN, +inf, and -inf.

Jan 31. To make the Assignment 1 more educational, you are NOT allowed to use the Matlab exponentiation function ^ in Problem 2.

Jan 31. I will pay THREE dollars per typo/mistake in the posted GVL chapters. Rules:

Jan 31. There are two typos in the original A1.pdf. In Problem 2 it should be m=1 if norm(A,1)<<=1 not m=0. In Problem 3 it should be "x(2:2:r)" not "x(2:r)". The A1.pdf now posted is correct.

Jan 28. The GVL chapters include way more material than we cover in the course. There are two ways to deduce the proper subset of material that you need to understand. (1) The assignments and lectures define what is very important. (2) The syllabus page identifies those sections and subsections that house required material.

1.	Typos must be emailed to cv@cs.cornell.edu. To receive payment, you must be the first to spot the mistake.You will receive a confirmation soon after submission.
2.	In general, there will be a maximum of one typo per equation. Thus, F = (9/50)C + 31 is worth one typo.Same rule for lines of pseudocode.
3.	Maximum of one typo per bibliographic entry.
4.	Payment at the end of the semester.
5.	I reserve the right to call the whole thing off at any time! If that happens, payment will be made at the time when this opportunity is cancelled.