structure_momentum{note, "3/7"}
structuring your code and analysis, in a way that allows for continuous improvement over a period, without getting into more and more of a tangled mess of impossible to understand code...
and having no idea what a copy was and why another one was better.
create separate notebooks for each attempted model, duplicating and editing, renaming and ordering carefully.
a low-tech approach ensures clarity between models experimentation.
it's simple yet effective for staying organized.
versioning and linking to git commits.
1 notebook, 1 submission (before the next model attempt).
(don't be french be italian "lol", so...)
- keep things simple where we can.
- don't overcomplicate things.
exp(erience/ertise):
for me, there's always a version that's closer to the final one i have in mind, or leans more towards the ideal.
if i don't have a clear, defined idea and i'm following intuition or inspiration...
or i'm just playing around, which always leads to great discoveries...
[personally: i delete outdated ones to travel light].
- exploratory programming in python is neat.
it's a great way to try things.
when done, just keep the final version. - trying to create their own architectures or variants.
key tips: 1 notebook per model/experiment (navigate orderly through versions)
- duplicate + rename notebooks for each project phase >> order.
- maintain logical file organization [debugging + progress] >> management
- clear/descriptive names for files/models [to know what you're looking at] >> clarity.
structure for iteration:
for rapid iteration, everything needs to be quick and easy.
slow and difficult things not only consume time but also mental energy.
structure your code and analyses to allow for continuous improvement over time, without ending up with an unmanageable tangle of incomprehensible code.
link notebooks to git commits to track versions.