InterFit.jl 🌌📡

An interactive tool for fitting interferometric models to Very Long Baseline Interferometry (VLBI) visibility data, with Bayesian parameter and uncertainty estimation.

Features ✨

Think of it like difmap's modelfit, but more automatic, with robust uncertainty quantification, and rich real-time diagnostics.

Core Capabilities

• Automatic Fitting 🎯: No manual component placement needed - finds globally optimal solutions using MCMC sampling
• Flexible Multi-Component Models 🔧: Flexible circular and elliptical Gaussian components powered by InterferometricModels.jl
• Bayesian Uncertainty Estimation 📊: Full parameter uncertainties via Pigeons.jl parallel tempering
• Flexible Data Handling ⚡: Fit to either complex visibilities or visibility amplitudes
• Real-Time Model Diagnostics 📈: Instantly see how well your model describes the data

Installation & Setup 🚀

1. Clone the Repository 📥: Download InterFit.jl to any location on your computer

   git clone https://github.com/aplavin/InterFit.jl.git
   cd InterFit.jl

   Alternative: If you don't have git installed, download the ZIP archive from GitHub and extract it.

2. Install Julia 💎: Ensure you have Julia 1.11 installed (installation guide)

3. Initialize the Environment 🔧: Start Julia with the project environment

   julia --project

4. Install Dependencies 📦: In the Julia REPL, run:

   julia> using Pkg; Pkg.instantiate()

Quick Start 🚀

Basic Usage

Load InterFit and start fitting models to your VLBI data:

# Load the package
julia> using InterFit

# Option 1: Direct file loading (recommended for most users)
julia> interfit("path/to/your/visibility_file.uvf")

# Option 2: Pre-processed data (for advanced users)
julia> uvtbl = load_average_uvtbl("path/to/your/file.uvf")
julia> # process/filter uvtbl as needed
julia> interfit(uvtbl)

# Option 3: Try with sample data (no file needed!)
julia> vis_fits = joinpath(pkgdir(VLBIFiles), "test", "data", "vis.fits")
julia> interfit(vis_fits)

Understanding the Interface 📊

InterFit opens an interactive window with three main areas:

Left Panel: Visibility Data vs Model Comparison

• Visibility plots: Amplitude and phase vs UV distance
• Residuals plot: Shows fit quality in units of measurement uncertainty (nσ)

Center Panel: Additional Diagnostics

• UV coverage: UV plane sampling of your observations
• Closure quantities: Phase and amplitude closures for visual model validation

Right Panel: Model Components

• Component positions: Sky locations and shapes of fitted Gaussian components
• Flux vs size: Component flux, angular size, brightness temperature
• Shape parameters: Major/minor axis sizes for elliptical components

Fitting Your Model 🔧

1. Data Preparation

• "Data σ adjustment": Fine-tune visibility weights
  • Multiplicative: Scale all uncertainties by a factor (default: 1.0)
  • Additive: Add systematic uncertainty floor (default: 0.0 Jy)

2. Choose Data Type

• "Fit to" controls what quantities to fit:
  • Vis: Complex visibilities (full information, assumes proper calibration)
  • Ampl fast: Visibility amplitudes with Gaussian likelihood
  • Ampl slow: Visibility amplitudes with Rician noise (more accurate, but slower)

3. Model Configuration

• "n Gaussians": Number of Gaussian components
• "n elliptic": How many should be elliptical

4. MCMC Settings ⚙️

• "n rounds": Number of tempering rounds. Rule of thumb: 3-6 for exploration, 10+ for final results.
  • Each increment doubles runtime
• "n chains": Number of parallel chains, 8-20 typically good.
  • Scales runtime linearly
• For more details, see the Pigeons.jl documentation

5. Fit & Save

• "Fit" button: Run model fitting with current settings
• "Save" button: Export fitted model to a file

Tips 💡

• Start simple: Begin with 1-2 components, increase if needed
• Check convergence: Run fit multiple times - if results vary significantly, increase n_rounds
• Check residuals: Look for systematic patterns indicating model inadequacy

Requirements 📋

• Julia 💎: Version 1.11 (recommended)
• Operating System 💻: Cross-platform (Windows, macOS, Linux)
• Dependencies 📦: Automatically managed via Project.toml

Built On 🏗️

InterFit.jl is heavily based on powerful and composable Julia packages:

• VLBIFiles.jl 📁: Reading visibility files
• InterferometricModels.jl 🌟: Source structure models and their visibility calculations
• Pigeons.jl 🔬: Parallel tempering MCMC algorithms
• AccessibleModels.jl ⚙️: Convenient interface for model parametrization and fitting
• Makie.jl & MakieExtra.jl 🎨: Plotting and visualization

Contributing 🤝

Code contributions, bug reports, and feature requests are welcome!