Skip to article frontmatterSkip to article content

In this section, we will host all of my notes for the deep dive sections. Each of these sections will have information that is relevant for the application tracks but they will be relatively agnostic to be used for general purposes. It can be thought of as the topics in a classic numerical analysis course but from a data-driven modeling perspective.


Mini-TOC

Numerical Analysis

  • Function Approximation
  • Differentiation
  • Integration
  • Numerical Linear Algebra

Parameterizations

  • Representation Learning
  • State Space Models

Learning

  • Optimization
  • Inference
  • Sensitivity Analysis

Algorithms

  • Gaussian Processes
  • State Space Models

Programming

  • GeoData

Numerical Analysis


Function Approximation


Differentiation

Taking Derivatives is arguably the most import component in data-driven learning. It will serve as a foundation for all subsequent topics and application surrounding learning. Gradients in general are the workhorse of data-driven methods. In addition, thinking about how we parameterize our models often involve thinking about gradients which stem from classical numerical analysis.


Integration


Numerical Linear Algebra


Parameterizations


Algorithms

Gaussian Processess

State Space Models