Software: Matlab, Python, pytorch and tensorflow libraries

Course Outline

Day 1. Introduction. Concept and notions of dynamic models (Bellman equation, Euler equation, Markov decision process, state and control variables, recursive equilibrium, non-stationary models, etc.). Computational methods for dynamic models (projection, perturbation and stochastic simulation).

Practice: Solving selected problems and illustrating computer codes.

Day 2. Large scale models including heterogeneous agent models and the curse of dimensionality.  Methods of computational economics for reducing the computational expense (sparse grids, ergodic set methods, non-product integration, quasi-Monte Carlo methods, endogenous grid method, parallelization, etc.)

Practice: Solving selected problems and illustrating computer codes.

Day 3: Introduction to machine learning methods in data science (supervised and unsupervised learning, logistic regression, dimensionality reduction, clustering, principal component analysis, stochastic gradient, etc.).

Practice: Solving selected problems and illustrating computer codes.

Day 4: Deep learning and other advanced ML classification methods (neural networks, decision trees, support vector machine, etc.).

Practice: Solving selected problems and illustrating computer codes.

Day 5: Deep learning methods for solving high dimensional economic models (deep reinforcement learning and its application to dynamic stochastic models including heterogeneous-agent new Keynesian model with incomplete markets, large scale models of central banking, etc.). Conclusion.

Practice: Solving selected problems and illustrating computer codes.