Introduction to Deep Learning · HIT

Week 5   Part II · Training Infrastructure

Loss Functions & Metrics

Task-appropriate losses (cross-entropy, MSE, BCE); metrics; the train and eval loop.

Curated, free, canonical references for this week: a course or lecture, a book chapter, a video, and an authoritative blog post or official tutorial. Each opens in a new tab.

Course
Stanford CS231n: Linear Classification (Softmax and SVM) cs231n.github.io

Derives softmax/cross-entropy and hinge losses with probabilistic and information-theoretic views.

Book
Understanding Deep Learning (Simon Prince), Chapter 5: Loss functions udlbook.github.io

Derives MSE, cross-entropy, and BCE from maximum likelihood (free PDF).

Video
StatQuest: Neural Networks Part 6, Cross Entropy youtube.com

A clear, intuitive explanation of cross-entropy as the classification loss.

Blog / Docs
scikit-learn User Guide: Metrics and scoring scikit-learn.org

Authoritative reference for accuracy, precision/recall, F1, ROC AUC, MSE, and R2.

← Back to the Week 5 lab

PreviousWeek 4: Data PipelinesNextWeek 6: Optimization