We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Jorn Mossel - Modeling Extreme Events with PyMC | PyData Global 2023
Modeling extreme events with PyMC: Jorn Mossel demonstrates Bayesian approach using open-source Python package, updating knowledge with new data and highlighting uncertainty quantification for accurate predictions.
- Modeling extreme events: Jorn Mossel talks about modeling extreme events using PyMC, an open-source Bayesian modeling package in Python.
- Gaussian processes: Uses Gaussian processes to model spatial relationships between weather stations and predict extreme events at individual stations.
- Return periods: Describes return periods, which are the time between extreme events. For example, the return period of 10 years means that an event of a certain magnitude can be expected to occur every 10 years.
- Uncertainty: Highlights the importance of quantifying uncertainty in modeling extreme events. Bayesian methods can provide uncertainty estimates and allow for updating knowledge as new data becomes available.
- Comparing Bayesian and non-Bayesian methods: Shows how Bayesian methods can be more effective for modeling complex problems, such as extreme events, by incorporating prior knowledge and updating as new data becomes available.
- PyMC’s capabilities: Discusses PyMC’s ability to model complex distributions, incorporate prior knowledge, and perform uncertainty quantification, making it a powerful tool for Bayesian model fitting.
- Practical example: Uses an example of extreme rainfall data to demonstrate how PyMC can be used to model and predict extreme events. Demonstrates the ability to make predictions at new locations not covered by existing data.
- Code snippet: Shows a few lines of code to fit a Bayesian model using PyMC and sample from the posterior distribution.
- Uncertainty quantification: Highlights the importance of uncertainty quantification in modeling extreme events. Demonates how PyMC can be used to quantify uncertainty and provide probability distributions for predictions.
- Comparison with non-Bayesian methods: Discusses the limitations of non-Bayesian methods for modeling extreme events and how Bayesian methods can provide more accurate and uncertainty-aware predictions.