Explore probabilistic numerics for inference with simulations in this Fields Institute seminar presented by Philipp Henning from Universität Tübingen. Delve into high-dimensional models, numerical methods, and probabilistic programming. Learn about algorithm implementation, mixed information sources, and optimization techniques. Gain insights on single forward pass and mini-batching approaches. Discover how these concepts apply to machine learning advances and applications through practical examples and observations.
Probabilistic Numerics for Inference with Simulations
Overview
Syllabus
Introduction
Highdimensional Models
What are Numerical Methods
The Problem
Simple Approach
Probabilistic Programming
Inner Loop
Algorithm Implementation
Mixed Information Sources
Single Forward Pass
Mini Batching
Optimization
Fancy Picture
Conclusion
Initial Observations
Taught by
Fields Institute
