Comrade

Comrade is a differentiable modular modeling framework for use with very long baseline interferometry. The goal is to allow the user to easily combine and modify a set of primitive models to construct complicated source structures. The benefit of this approach is that is straightforward to construct different source models out of these primitives. Namely, a end-user does not have to create a separate source "model" every time they change the model specification. Additionally, most models currently implemented are differentiable with at least ForwardDiff. This allows for gradient accelerated optimization, and sampling (e.g. HMC) to be used with little effort by the end user.

Comrade does not currently have a native optimization or sampling interface. The reasoning for this is that different problems are amenable to different optimizers. Rather than including all optimizers in Comrade, expanding the number of dependencies, Comrade tries to make moving from an image model to objective function easy. As an example of this we To use perform inferences on data you can then hook into the vast array of different modeling and optimization packages in Julia. There are some small examples packages defining these interface such as ComradeSoss.jl which combines Comrade with Soss a probabilistic programming language. Other interfaces to e.g. Turing, BAT are planned.

Requirements

The minimum Julia version we require is 1.6, which is the current LTS release. In the future we may increase this as Julia advances.

• Comrade
• • Requirements
• Getting Started
• • Introduction to the VLBI Imaging Problem
• Model Interface
• API