Deepening your knowledge#
Completing the basic assignments that demonstrate you can apply what we cover in class to your own data, is enough for a B, to earn an A you need to extend your knowledge.
You have some different pathways to earning an A.
Extending Assignments#
Starting in week 3 it is recommended that you spend some time each week working on extensions to earn level 3 on the skills.
Use the feedback you get on assignments to inspire your extensions.
To submit these, submit the work to a separate <assignment>extended
branch so for assignment 2 extension, submit to assignment2extended
.
You should not extend every assignment since skills overlap and relate to one another.
I recommend making an issue that is a plan and asking for feedback before you work on extensions, so that you do not do too much extra work.
Some ideas:
extend A5 to add in a database source or combine the data in new ways, plus do extended EDA to earn level 3 for: access, construct, prepare, visualize and summarize
extend one of A7-9 to have experiments that more carefully analyze and compare different models at a task for one of the ml tasks (classification, regression, clustering) and visualize and summarize
Deployment and Distribution#
Warning
this is currently a draft, will be updated by 10/24
Instead of earning all 15 level 3s you can earn any 10 plus the following:
transform your portfolio to a publish-able portfolio by making it a jupyter book (mostly set up already, just need to clean up)
share a model and use a classmate’s model with huggingface (making the model card would count for process and fitting it could count for one of the 3 tasks)
if you go this route, I recommend level 3 for:
access (loading models from huggingface will count)
python
process (see Process Level 3)
summarize
visualize
evaluate
one of: classification, regression, clustering
optimize
compare
workflow
If your dataset of interest is hard to work with, images, or text, you might swap in representation instead of access
this means skipping level 3 for:
prepare
construct
2 of: classification, regression, clustering
representation