11. Course feedback¶
This homework is to be done individually, not in groups. Your responses are not to be put in your GitHub account so you can speak candidly without the influence of your teammates. You should not use a Jupyter notebook to do the problem. Instead, you should submit your responses either as a PDF or just as text in an email. Do not submit a MS Word document or a Jupyter notebook. Please email your responses to all of the following email addresses.
bois at caltech dot edu rfu at caltech dot edu troeschi at caltech dot edu atribby at caltech dot edu jwagner2 at caltech dot edu
A total of 20 points will be awarded for thoughtful responses for this homework. I am not expecting a paragraph for each question, but if you have detailed comments on something you have a strong or insightful opinion about, I would like to hear them.
Problem 10.1: Topical coverage¶
a) Are there topics you would like covered that were not?
b) Were there topics you think we spent too much time on?
c) Which topics did you find most interesting?
d) Which did you find most pertinent?
e) Where there any common misconceptions that persisted throughout the course for you? What topics (if any) did you find particularly hard to understand and apply?
Problem 10.2: Data sets and exercises¶
a) What was your favorite homework problem or data set? Which problem would you remove? Why?
b) Some of the homeworks involved slamming up against some computational roadblocks, from tough optimization problems to Stan sampling. In other parts of the homeworks, I gave you tips and watchouts. You will hit those roadblocks in the future, and my hope is that having encountered them in class, they will not be so bothersome. How do you feel about this amount of admittedly painful challenges? Should I have given more watchouts and build more of the code structure for you, should I have given you less tips and let you hit more roadblocks, or was it about right?
c) On the other hand, I worked to abstract away too much grungy work. Most of the
bebi103 package is meant to help you avoid doing parsing and be able to write declarative code. How do you feel about this? Should I have facilitated more or less, or what is about right?
Problem 10.3: Course structure¶
a) Where did the majority of your learning take place? (Example answers include: Reading the lessons, doing the lesson exercises, completing the homeworks, interacting with TAs during the Tuesday sessions, interacting with TAs during office hours, during JB’s lectures, during JB’s office hours, talking with teammates, etc.)
b) Did you attend any recitation sessions? Did you find them useful? Do you have general suggestions for that?
c) Do you have any comments on the lecture style or material?
d) Did you read any of the suggested readings in the Readings/tutorials section of the course website? If so, which did you find particularly useful?
Problem 10.4: Your future¶
a) What will you be able to do with your data that you weren’t able to before?
b) Has this class changed the way you read scientific papers? Has it changed your opinion on any papers you’ve read previously?
c) Which of the following principles/techniques do you think you will apply in your work going forward?
Bayesian generative modeling in general
Brute force plotting of posterior distributions
MAP finding and reporting credible regions making a local Normal approximation
Markov chain Monte Carlo
Posterior predictive checks
Model comparison using WAIC and LOO
Principled pipelines including prior predictive checks, ground truth checking, shrinkage checking, and simulation based calibration
Nonparametric Bayes, and Gaussian processes in particular
d) Which of the following software tools do you think you will use in your work going forward?
Python-based tools in general (NumPy, SciPy, etc.)
e) Will you share what you have learned with your labmates/classmates?
f) Will you share what you have learned with your PI?