AI Machine Learning Basics: Understand Neural Networks Through Games

🧮 Machine Learning Basics — Understand Neural Networks Through Games

🎮 Analogy: ML Is Like Training a Pet

Train a Pet	Machine Learning
Treats reward correct behavior	Loss function + backpropagation
Repeat to form a habit	Multiple training iterations (Epochs)
Gradually reduce mistakes	Loss drops
Finally learns a new skill	Model converges, accuracy met

🧱 Three Layers

• Input layer: receives raw data (e.g. 784 pixels)
• Hidden layer: feature extraction and transformation
• Output layer: outputs final result (e.g. 10-class probabilities)

🎯 Hands-on: Digit Recognition Model

1. Load MNIST handwritten-digit dataset (60,000 images)
2. Design network: 784 input → 128 hidden → 10 output
3. Choose ReLU + Adam optimizer
4. Click train and watch the loss curve drop
5. Test: can it recognize your handwritten digit?

📊 Key Concepts

• Epoch: number of passes over all data
• Accuracy: proportion predicted correctly
• Overfitting: memorized answers but can't generalize