A compilation of wisdom from Professor Isbell and others. This was compiled and posted on Piazza at the end of the course (not by me)
See also: http://cs.brown.edu/~mlittman/etc/charles.txt
Q: If the submission was accepted it is not considered late?
A: Please stop asking us if it’s late, and explaining to us how the internet works... The reason there’s a buffer is so we don’t get all of these posts and emails explaining how you submitted it on time but, man oh man, my internet was really slow just tonight and etc etc etc which for some reason always seems to happen the night an assignment is due.
Q: Personally, I did not know the due time is actually 9:15pm PST until 9:00 pm. It is a negligence on my part, but please also understand that this is an online course and miscommunication and misunderstanding do happen.
A: I have my own deadlines that involve folks with no sense of humor whatsoever, so for the sake of efficiency, doesn't "It is a negligence on my part" end this entire conversation? I think it does.
Q: Is it a requirement to regular post/respond to posts in Piazza?
A: Why in the name of all that is holy would I want such a thing? I will say this much: I’m going to start docking points for anyone who uses the word “confirm” in any context on any piazza post. Sort of anti-participation points.
Q: Can we use any dataset(s) as long as they fit the scope of the assignment?
A: You’ve constructed a question with its own answer in it. It’s poetically tautological. How would you change the question so that the answer could be no?
Q: Book is poorly written imo (overly verbose / not concise)
A: The irony question was around you saying “overly verbose / not concise” which is twice as long as it had to be.
Q: In my DT Classifier I am using gini index. is it okay? In the write-up here is what it says: "You are not required to use information gain (for example, there is something called the GINI index that is sometimes used) to split attributes, but you should describe whatever it is that you do use."
A: I mean... the quote you provide literally answers your question. Do you not believe it?
Q: OK, there are no negative rewards here as we can't get negative grades
A: Who says you can't get negative grades?
Q: For boosting we choose a decision tree as a weak learner because we know it will do better than chance at every time step. I was wondering why this would always be true?
A: “We know it will do better than chance” Do we? XOR would like a word with you.
Q: I understand how to find the optimal number of components for PCA but it's not clear how to do this for ICA.
A: Wait. You understand how to find the optimal number of components for PCA? Are you sure?
Q: Heh, its just posting a link, I suspect there is a different reason for holding off.
A: “It’s just posting a link” is like saying “it’s just physics”.
Q: Where is the reading Mitchell? I looked through the Files in canvas and wasn't sure what it was called
A: It's the book
Q: Is that the jist of this, beyond the other specifics about submission format and tools and blah blah blah?
A: “Gist”
Q: Because MIMIC uses less evaluations and structure this might be an interesting case.
A: “Fewer”
Q: Is content not covered in the lectures, but covered in the indicated chapters in the book still fair game to appear on the midterm?
A: In your heart you already know the answer to this question, don’t you (I mean beside the fact I’ve answered this elsewhere)?
Q: What are covered in midterm?
A: This has been covered 8000 times. And it is pinned because it’s in the FAQ
Q: Can we use calculator in the midterm?
A: Have you read the FAQ?
Q: I may not be able to take it in time before I leave so I'll try to find a room tonight that's quite or I'll hope I don't break down on the road on Sunday.
A: You mean emotionally?
Q: Can we get a heads up when word is received that it's no longer crazy so I can log on?
A: Well if it’s crazy you won’t be able to log on.
Q: We aren't ever getting that video are we?
A: I’m sure you will. I mean if people stop asking me about it so I can do stuff.
Q: Still no grade yet...
A: It's weird. I've been staring at this pot for awhile and the water hasn't boiled yet. (For those not from around where I'm from, there is an expression that applies here: "a watched pot never boils").
Q: Do you have an idea about the percentage of the grade to get A , B and C ?
A: Yes
Q: When asked to implement 'Decision trees with some form of pruning' can we implement random decision forest?
A: Porque no los dos?
Q: Can course staff clarify which Schapire introduction paper is the right one?
A: Porque no los dos?
Q: I assume that the 12 page limit is ALL inclusive (graphs, figures, appendixes, references...etc.)?
A: Nothing past 12 pages will be read. As for the rest, don’t do anything dumb. 9 pt font and .25” margins and I will destroy your credit rating.
Q: How many lectures should be watched to complete HW1?
A: This question seems like the kind of question that answers itself as you watch the lectures.
Q: Confusion matrix can be added to report? In addition to graph can that be used for explaining analysis?
A: Why would the answer to this sort of question ever be no? Are you imagining that you’d lose points with an explanation like, “how dare you provide more support for you analysis! -10. No, -15, you scum.”?
Q: What language is everyone using for assignment 2?
A: LISP. I’m pretty sure it’s a requirement.
Q: Can we use multiple languages for Assignment 1?
A1: I'm pretty sure the answer is, "no me importa".
A2: So let me see if I understand the question. You are imagining that I’m the kind of person who doesn’t care at all what programming languages you use, what packages you use, or even whether you completely steal the code or packages, but that I would care very much whether you steal all this code from the same place. Is that a decent summary of the premise behind the question?
Q: Can some instructor/TA respond if these lectures are necessary or not for the first assignment?
A: I suspect you’ll be able to answer that once you’ve looked at the material and the assignments.
Q: Preprocessed dataset good for subsequent assignments?
A: Try reading the text for assignment #3 and see what you think.
Q: So I am wondering if using Weka for assignment is allowed since it is so...
A: Re-read your question carefully. In it you will find your own answer. Also you might want to re-read the assignment carefully.
Q: Would it be permissible to construct a dataset composed of pre-trained convnet (e.g. ResNet50) features from an image dataset?
A: You already know the answer to this question.
Q: Does it matter what version of python we use?
A: You already know the answer to this question in your heart.
Q: Can we use datasets from openml.org?
A: I’m completely sure you already know the answer to this question.
Q: On at least one occasion a professor told us we could use whatever language we wanted, so I picked Ruby. (This was circa 2005.) As a result, on the next assignment we were told we had to pick Java or C. To his credit he graded it gamely.
A: Yes, but the difference is I don't care. I'm not trying to be nice to you people—I assume you're all bots—I just don't care.
Q: Can we reuse Decision Tree / Bagging-Boosting implementation from ML4T?
A: I don’t care. At all.
Q: Just to make sure... we can use sci-kit learn / keras for these implementation ?
A: I don’t care. No matter how many times you ask me a version of this question, I will say I do not care.
Q: For the a1, it mentioned that we should be responsible for the code that can run on CoC machine, I am wondering whether we can include package install command line for R script
A: I so don’t care. I so so so don’t care. Use your README appropriately.
Q: Shall we use python 2.6.6 to do the assignment 1?
A: Internalize that I don’t care about these things. I just do not care. </believe me when I say these things>
Q: Can it be assumed that the CoC boxes can be updated to run a version of python that is sane to use in 2018?
A: Read my instructor’s response if you haven’t. If you have, carefully re-read it again.
Q: That said, I’m getting the impression that the professor here genuinely doesn’t care. Mostly because he’s said many times that he doesn’t care.
A: Have I mentioned recently that I don't care?
Q: So there's a github submission on MIMIC in python; it seems to be maintained by a former OMSCS student. As long as we're citing properly, we should be able to use that code, right?
A: The correct answer, of course, is “We still don’t care. OMG NOTHING MATTERS.’
Q: Can I use Jython for the assignment 2?
A: I can’t believe you’re asking me this question.
Q: Can I use python MDP to finish assignment 4?
A: I am not at all amazed this kind of question is still being asked.
Q: So do you have a preference whether we use tool X or programming language Y or library Z to complete the assignment?
A: No.
Nope.
Nay.
Nyet.
None.
Nil.
Zero.
Zip.
Zilch.
Nada.
Nicht.
Nein.
Niet.
Nie.
(man falling from the rooftop) N-n-o-o-oooooooooo.
...
...
No.
Q: I take that as an implicit yea
A: Every single question of this forum has an implicit yea
Q: I think the answer to every such question is yes :) It can get as complicated as we want, that is my understanding.
A: Indeed.
Some of these responses are hilarious, but I'll probably change this opinion once I actually start taking the class... Haha