Recent advances in artificial intelligence and rapidly expanding computational power have provided economists with unprecedented capabilities for numerical analysis. This course offers an overview of numerical methods at the intersection of mathematics, statistics, and computer science, essential for modern economic dynamics. It is divided into three parts: Part I introduces foundational tools of numerical analysis, including approximation, integration, optimization, and error analysis, together with both local and global solution techniques. Part II explores computational methods tailored to high-dimensional problems, such as Smolyak and sparse grids, derivative-free solvers, low-discrepancy sequences, endogenous grids, and epsilon-distinguishable sets. Part III covers machine learning approaches, including supervised and unsupervised learning, deep learning, reinforcement learning, decision trees, support vector machines, parallel computing, and big data methodologies. Applications include economic models - new Keynesian, default risk, heterogeneous agents, international trade, and growth models - and computer science examples, such as handwriting recognition. Programming uses Python and MATLAB. Assessment is based on problem sets and a final project.
2-5 units · Letter or Credit/No Credit
Recent advances in artificial intelligence and rapidly expanding computational power have provided economists with unprecedented capabilities for numerical analysis. This course offers an overview of numerical methods at the intersection of mathematics, statistics, and computer science, essential for modern economic dynamics. It is divided into three parts: Part I introduces foundational tools of numerical analysis, including approximation, integration, optimization, and error analysis, together with both local and global solution techniques. Part II explores computational methods tailored to high-dimensional problems, such as Smolyak and sparse grids, derivative-free solvers, low-discrepancy sequences, endogenous grids, and epsilon-distinguishable sets. Part III covers machine learning approaches, including supervised and unsupervised learning, deep learning, reinforcement learning, decision trees, support vector machines, parallel computing, and big data methodologies. Applications include economic models - new Keynesian, default risk, heterogeneous agents, international trade, and growth models - and computer science examples, such as handwriting recognition. Programming uses Python and MATLAB. Assessment is based on problem sets and a final project.
Offered in Autumn 2025 at Stanford University.