11-785 Introduction to Deep Learning
Spring 2020

Piazza Etiquette

Asking questions?

When asking questions, first check if someone else has asked a similar question. If they have, we are simply going to redirect you to that answer anyway.

If you have a new question (or iteration on a previous question), please try to make the title as descriptive as possible so that others can search for your question. For example, if you are asking about HW0 question 3.3, make sure those words appear in the title. And if you are going to ask a question about HW0 question 3.3, try searching that up on Piazza before you ask.

Your questions should be clear. We should be able to understand what your question is asking and your situation should be laid out such that we are able to help you.

Private vs. Public

Only make a post private if you have a problem that is specific to you. I.e. you need to share your code to get help debugging or anything else extremely personalized.

Public posts will help everyone learn, but if there are too many public posts on Piazza, then Piazza will become useless and messy. So, be entirely sure that your question cannot be answered with a quick Google Search. Before asking your question, think about how you can frame it such that it can benefit you and other students in the course.

Good Questions vs. Bad Questions

With over 400 students (as of writing) and more Piazza questions already asked than some courses at CMU over the entire semester, we have a challenge at hand (giving extremely fast and beneficial support)- one we hope to live up to.

Good questions will help us achieve our Piazza goals. Here is an example of a really good question as follows. It is numbered so we can easily respond to individual questions, and it references the recitation number in the title. Additionally, it shows us that we may need to do a better job in the future of explaining something.

  1. What does parameters of subnetworks will be “absorbed” into bigger networks mean?
  2. On page75 of the slides, there is nn.ModuleList(self.hidden); in the notebook example, there is nn.Sequential(*layers). What does it mean to use the asterisk(*)?
  3. Why gradients accumulate in Pytorch while it must be set to zero each time? Are there specific use cases for this kind of setting?

Good questions will inherently show us that you have thought about the problem you are asking, have a clear idea of the problem, and have tried to debug the problem on your own. Bad questions exist and will hurt our ability to answer you in an efficient manner.

Always feel free to ask us to elaborate on a response if needed, and try to be as polite as possible.