IN1160 — Introduction to ML & AI

IN1160 – Introduction to Machine Learning and Artificial Intelligence at UiO. A practical and theoretical introduction to the core ideas behind modern AI.

Course topics

Representation and vector spaces

• Vector Space Models — representing text as vectors, TF-IDF weighting, cosine similarity
• Document classification using vector representations
• Dimensionality and sparsity in NLP

Classical machine learning

• Regression — linear and logistic, gradient descent, regularisation (L1/L2)
• Classification — k-NN, naive Bayes, SVM
• Decision trees and random forests — information gain, Gini impurity, ensemble methods
• Evaluation metrics — accuracy, precision, recall, F1, confusion matrices, cross-validation

Neural networks

• Feedforward networks: layers, weights, activation functions
• Backpropagation and the chain rule
• Training: optimisers (SGD, Adam), learning rate scheduling, early stopping
• Preventing overfitting: dropout, regularisation, batch normalisation

Unsupervised learning

• K-Means clustering and the EM algorithm
• Dimensionality reduction: PCA and t-SNE for visualisation

Reinforcement learning

• Markov Decision Processes (MDPs): states, actions, rewards, policy
• Q-learning: tabular and deep (DQN) variants
• Exploration vs exploitation trade-off

Generative AI

• Generative models overview: VAEs, GANs, diffusion models
• Language models: n-gram models to transformer attention (conceptual)
• Practical use of large language models: prompt engineering, limitations

AI ethics

• Bias in training data and model outputs
• Fairness metrics: demographic parity, equalized odds
• Transparency, explainability, and accountability
• Case studies: facial recognition bias, hiring algorithms, medical AI

Takeaway

AI is not magic — it’s applied statistics and optimization. Understanding the math (gradient descent as loss surface navigation, attention as weighted query-key similarity) makes it possible to reason about why a model fails, not just accept that it does.