2 Comments
User's avatar
Neural Foundry's avatar

Really thoughtfull reconstruction of the CS231n material. Your Wally and Louise analogy for understanding loss functoins actually clarifies something most textbooks gloss over, which is that loss isn't just a metric but a deliberate inversion ofhow we naturally think about correctness. The regularization section is paticularly valuable because it connects the mathematical formalism to the practical problem of overfitting in a way that makes the penalty term feel less arbitrary.

Expand full comment
Ajay Kapal's avatar

These topics are foundational, but they're conceptually tricky especially the first time you see them. Writing helps me understand, and I'm glad it helped you. Thanks!

Expand full comment