Opening Harnessing Chaos and Predicting the Unpredictable with Artificial Intelligence
Harnessing Chaos and Predicting the Unpredictable with Artificial Intelligence
- Length
- 1 h 20 min
- Level
- Undergraduate
- Recorded
- Santa Fe Institute2018
Chapter 1 of 10: Before the lecture
Trains only the output layer of a reservoir computer to forecast toy weather models past the Lyapunov time.
Reservoir computing, a recurrent neural network whose internal dynamics stay fixed while only its output weights are trained, can forecast chaotic systems such as weather. Girvan, a University of Maryland physicist, starts from the butterfly effect and the Lyapunov time, then contrasts this knowledge-free approach with deep learning. Toy models of spatiotemporal chaos serve as test cases, and parallel reservoirs split large systems into local neighborhoods without extra training time. A detour applies reservoir computing and causal state models to user activity on Twitter. The Santa Fe Institute Community Lecture ends with a hybrid of knowledge-based physics models and reservoirs, then machine ensembles and audience questions.
Reservoir computing, a recurrent neural network whose internal dynamics stay fixed while only its output weights are trained, can forecast chaotic systems such as weather. Girvan, a University of Maryland physicist, starts from the butterfly effect and the Lyapunov time, then contrasts this knowledge-free approach with deep learning. Toy models of spatiotemporal chaos serve as test cases, and parallel reservoirs split large systems into local neighborhoods without extra training time. A detour applies reservoir computing and causal state models to user activity on Twitter. The Santa Fe Institute Community Lecture ends with a hybrid of knowledge-based physics models and reservoirs, then machine ensembles and audience questions.
Phys. Rev. Lett. (2018)
International Conference on Social Computing (2013)
Predictability of User Behavior in Social Media: Bottom-Up vs. Top-Down Modeling

This lecture
Talk 24 of 24 of Foundations of Complex Systems.
Choose a node to see how it connects to this lecture. What comes before is on the leftat the top. The ring is what it teaches. People and places are above. Other lectures and books are below.
Press a chosen node again to open it.
Your comment is kept while you sign up.Sign in
Hybrid forecasting of chaotic processes: using machine learning in conjunction with a knowledge-based model
The Wisdom of Crowds