Loss Functions: Why MSE and Cross-Entropy Shape How AI Learns

The loss function is the single number that tells an AI model how wrong it is. Choose the wrong one and your model fails — even if the architecture is perfect. In this episode, we compare the main loss functions: Mean Squared Error (MSE), Mean Absolute Error (MAE), Binary Cross-Entropy, and Categorical Cross-Entropy. When to use each, why MSE is bad for classification, and how the loss function shapes what your model actually learns.

🎯 CHAPTERS
00:00 Why the loss function matters more than you think
1:00 Mean Squared Error (MSE) — regression classic
3:00 The outlier problem → Mean Absolute Error (MAE)
5:00 When the target is a probability (classification)
6:30 Binary cross-entropy: the formula explained
9:00 Why NOT to use MSE for classification
11:00 Categorical cross-entropy (multi-class)
13:00 Choosing the right loss function
14:30 How loss shapes what models learn

📌 KEY CONCEPTS
• MSE: L = (1/n) Σ (y − ŷ)²
• MAE: robust to outliers, L = (1/n) Σ |y − ŷ|
• Binary cross-entropy: L = −[y log ŷ + (1−y) log(1−ŷ)]
• Categorical cross-entropy: multi-class extension
• Why MSE gradients vanish in classification
• Loss function design = model behavior

📚 PREREQUISITES
• Episode 1 — Linear Regression: youtu.be/...
• Episode 4 — Backpropagation: youtu.be/...

📚 THE SERIES
Episode 5 of 10. Next: Activation Functions — why ReLU killed the sigmoid, and why GELU dominates modern transformers.

📱 WANT TO PRACTICE THE MATH?
NovaMaths — SAT & ACT Math prep with 749+ exercises:
https://www.novamaths.app

🌍 French channel: ‪@MathsAcademy27‬

─────────────────────

Channel hosted by Julien, certified math teacher with 30 years of classroom experience.

#LossFunction #CrossEntropy #MSE #Classification #AIMath

Видео Loss Functions: Why MSE and Cross-Entropy Shape How AI Learns канала Math Vision