Course

This is the recommended order for turning the atlas into a clean AI course.

1. Foundations

Start with concepts before model names.

• Core concepts
• Linear Regression
• Logistic Regression

2. Local Rules and Similarity

These lessons build intuition for decision boundaries, feature space, and nonlinear behavior.

• KNN
• Decision Tree

3. Ensembles

Move from one unstable tree to averaged or boosted collections of trees.

• Random Forest
• Gradient Boosting

4. Unsupervised Learning

Introduce structure discovery without labels.

• K-Means
• PCA

5. Deep Learning

Deep learning uses multi-layer neural networks to learn useful representations directly from data. Study it after the classical baselines so you can tell when additional model capacity is actually useful.

Before the architectures, revisit the concepts that make neural-network training understandable:

• Loss and Optimization
• Regularization
• Generalization

5.1 Neural building blocks

Begin with one artificial neuron, then follow the error signal backward through a tiny network.

• The Perceptron
• Backpropagation

5.2 Feed-forward networks

Start with the smallest general neural network. Focus on dense layers, nonlinear activations, backpropagation, and the difference between parameters and hyperparameters.

• MLP

5.3 Sequence architectures

Then compare two ways to model ordered data: causal convolutions for controlled local context, and attention for flexible global context.

• TCN
• Transformer

By the end of this section, you should be able to explain representation learning, a training loop, overfitting, regularization, and why deep learning is powerful without treating it as the default for every dataset.

6. Generative Models

End with modern generative modeling once the learner has a working vocabulary for optimization, representation, and evaluation.

• Diffusion Model

7. Browse the Atlas

Close the loop by revisiting models through different goals, data types, mechanisms, and constraints.

• Categories