Encyclopedia · 176 concepts

Math & Optimization · advanced · concept 50 of 176

Monte Carlo Methods

Answering hard questions by random sampling: estimate what you cannot compute by simulating it many times. The name behind MCMC in Bayesian inference, Monte Carlo returns in reinforcement learning, dropout-based uncertainty, and the tree search that powered AlphaGo. One idea, remarkable reach.

Key terms

Random samplingMCMCMonte Carlo Tree SearchSimulationVariance reduction

Where you meet it in the real world

AlphaGo's tree search, risk simulation in finance, Bayesian posterior sampling, uncertainty estimates

Videos

6. Monte Carlo Simulation

MIT OpenCourseWare · YouTube