Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Attend this 57-minute seminar to explore how Agent-Based Models (ABMs) can be enhanced through data-driven approaches and machine learning techniques. Discover how to transform traditional ABMs into probabilistic generative models that can learn latent parameters and micro-level variables from real-world data. Learn about the key design principles of balancing stochasticity with data availability and replacing unobservable discrete choices with differentiable approximations. Examine how maximum likelihood estimation and gradient-based optimization can improve parameter calibration compared to simulation-based approaches. Explore practical applications including opinion dynamics models applied to social media data and variational inference methods that bypass explicit likelihood derivation. Gain insights into the importance of interpretable scientific models in an era dominated by black-box deep learning systems. The presentation covers computational social science applications, data mining techniques for complex systems, and the intersection of traditional modeling with modern machine learning approaches. Understand how this methodology addresses common limitations of ABMs including limited predictive power, lack of principled parameter calibration, and inability to estimate agent-specific state variables.
