Problem 44508. Curve fitting (non-linear functions) & function handles
I think you may need to modify some of the tests to "...ensure both odd and even values of x" in all cases where m might be negative.
Hi, Tim. Thanks for your feedback. Actually, from the Player's point of view, m 'could' be negative in any Test. I agree that in some situations there might not be one unique pair of the m & c parameter values that is correct. However, I never apply an assertion to the values of m & c in the Test Suite, rather I check what values of y the user-supplied parameters produce from the user-supplied function(handle). _Any_ valid combination of m & c should be able to REpredict the same values of y provided in the original input, from the same x values (or a subset thereof). There is only one test where I check for prediction of y using x values not included in the original input (extrapolation/interpolation), and for that one test I do have to be careful to ensure there is one unique pair of m & c values that is correct. So I would say this is intentional (I'll add a short note to the Problem Statement). But please let me know if there's a flaw in my logic. —DIV
Aha--I see now. To quote Emily Litella: "Never mind."
I should clarify one detail for other Players. Given y = mˣ + c (elementwise), suppose x is the vector [7 8 5 6] and y is the vector [123 123 123 123]. Then there are two valid combinations of m & c, namely m=0 & c=123 and m=1 & c=122. So actually there is not just one unique set of m & c values in this case. However, given x>0 (in the Problem Statement), either of these alternatives can be correctly extrapolated or interpolated from.
This solution fails sometimes, and I have no idea why (it's just a "smart" brute force code). I've tried to reproduce the error several times on my pc, but I wasn't able to. And I've even tried to use python's long to no avail . Probably I am underfitting at some cases, but it works sometimes. :)
There is definitely some issue.
From my perspective in writing the Test Suite a solution should either pass all of the Tests every time, or else fail at least one test every time.
I have therefore increased the number of testing iterations in selected Tests, and your solution consistently fails the revamped Tests 4, 5, and 6; it may also fail some other Tests that haven't been amended (e.g. Solution 3651713 failed test 7a).
I have also added quite a bit of detail in the rejigged failure messages so that you can identify problematic parameter values.
Problem Recent Solvers2
Make an awesome ramp for a tiny motorcycle stuntman
USC Spring 2012 ACM: Armageddon
More from this Author32
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!Start Hunting!