Checklist

Machine Learning study checklist

A machine learning checklist that bridges math to code, demystifies hyperparameters, and builds a paper-reading habit through project sprints.

Built for Machine Learning · Students balancing math foundations with project implementation.

Progress

0 of 13 tasks complete

Solidify the math foundations

The linear algebra and calculus ML rests on.

Review linear algebra for MLMatrix multiplication, eigenvectors, and projections underpin most models.~55 minReview gradients and the chain ruleBackpropagation is just the chain rule applied repeatedly.~50 minConnect probability to loss functionsSee cross-entropy and MSE as likelihood-based objectives.~45 min

Translate math to code

Turn equations into working models.

Implement linear regression from scratchCode the gradient-descent update from the loss equation without a library.~60 minImplement logistic regressionAdd the sigmoid and cross-entropy loss; verify gradients numerically.~60 minVectorize the implementationsReplace loops with matrix operations and confirm identical output.~50 minBuild a training and evaluation loopTrack train and validation loss to spot overfitting early.~50 min

Tune and evaluate

Understand what each knob does.

Map hyperparameters to behaviorLearn how learning rate, regularization, and batch size each change training.~55 minRun controlled experimentsChange one hyperparameter at a time and log the effect on validation loss.~55 minUse proper validationHold out a test set and use cross-validation to avoid leakage.~45 min

Read papers and ship a project

Apply learning to real work.

Read one paper with the three-pass methodSkim, then read for ideas, then read for detail; summarize the contribution.~60 minRun a project sprintTake a dataset from baseline to an improved model with documented experiments.~90 minWrite a results summaryReport metrics, what worked, and what you would try next.~40 min

Common mistakes

Reaching for a library before understanding the math the model implements.
Tuning several hyperparameters at once so no effect is interpretable.
Evaluating on the training set and mistaking memorization for learning.
Reading papers linearly and getting lost instead of using a three-pass approach.
Skipping numerical gradient checks and shipping a broken backprop implementation.

Pro tips

Implement each model from the loss equation before using a framework.
Change one hyperparameter at a time and log its effect on validation loss.
Always hold out a test set and watch the train-validation gap for overfitting.
Read papers in three passes: skim, ideas, then full detail.
Verify gradients numerically before trusting a from-scratch implementation.

FAQ

How should I start the Machine Learning study checklist?

Start with the first phase, then run one timed Study Spaces sprint before adding more tasks. The goal is execution, not a perfect plan.

What should I do if I fall behind?

Copy the remaining tasks, pick the highest-score or highest-deadline item, and restart with one focused block.

How often should I review progress?

Review after each sprint and once at the end of the week so the next session starts with a clear first task.

Browse resources

Use it now

Turn this page into a live sprint

Start the matching room for Machine Learning, then use the sprint plan as the first task and recap script.

Machine Learning study checklist
Focus target: Machine Learning
Block 1 (25 min): closed-book recall or one timed practice set.
Break (5 min): mark confusing items without opening a new task.
Block 2 (25 min): correct misses and write the next first step.
Done: one score/error note plus one queued task for tomorrow.

Start Machine Learning roomResource list Compare methods