We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Charles D Lindsey - Bayesian Statistics with Python No Resampling Necessary | SciPy 2023
Explore Bayesian statistics with Python, a powerful regularization technique that can handle non-convex problems and offer flexibility in model design, through the lens of SciPy 2023 speaker Charles D Lindsey.
- Bayesian statistics is a type of regularization that can be seen as a subset of L2 regularization.
- There are many different types of regularization penalties that can be incorporated into a Bayesian model.
- Non-convex problems are still relevant and can be handled with Bayesian methods.
- Regularity conditions must be met for Bayesian methods to be valid.
- The flexibility of Bayesian models and incorporation of different regularization penalties is helpful in many cases.
- The limited memory BFGS bounded method is a good option for solving non-convex problems, but may not work well for very non-convex problems.
- Hyperparameters can be used to adjust the strength of regularization in a Bayesian model.
- Bayesian methods can be used for problems that are convex, non-convex, or mixture of both.
- The Hessian matrix can be used to compute variance in a Bayesian model.