Mathematical foundations for AI
- Fundamentals of linear algebra: conceptual understanding and simple calculations, such as matrix multiplication and vector inner products.
- Fundamentals of probability and statistics, such as mean and variance.
Basic coding
- Read and trace simple Python code. Writing Python code is not required.
- Read pseudocode and explain its logic.
Machine learning
- Supervised learning (e.g., linear regression, logistic regression, support vector machines, decision trees, kNN, ensemble learning, bias-variance tradeoff, cross-validation, and loss functions).
- Unsupervised learning (e.g., k-means clustering and principal component analysis).
Deep learning
- Multilayer perceptron models
- Transformers
- Large language models
- Computer vision
- Generative AI
What students should understand
Most problems are designed to be solved by reasoning from first principles. Focus on the core ideas behind AI methods, why they work, and what their outputs mean.
Quantitative questions require basic math and simple calculations, such as computing a dot product. Students should understand why a dot product can help measure similarity and how a vector can represent an image or a sentence.
Students may be asked to read simple Python and pseudocode. The exam does not require writing code, memorizing NumPy, pandas, or scikit-learn APIs, or reading NumPy or PyTorch code.