FURAX-CS Documentation

GPU-accelerated CMB parametric component separation built on JAX and FURAX.

FURAX-CS implements adaptive K-means and multi-resolution clustering strategies for spatially-varying foreground modeling of Cosmic Microwave Background (CMB) data, with benchmarks against FGBuster.

Installation

# Install JAX (CPU or CUDA)
pip install -U "jax[cpu]"       # CPU-only
pip install -U "jax[cuda]"      # GPU (recommended for production)

# Install furax-cs
pip install -e ".[all]"

User Guide

• Quick Start (Python API)
  • Data Loading
  • Running Component Separation Programmatically
• CLI Reference
  • Workflow Overview
  • generate_data
  • kmeans-model
  • ptep-model
  • fgbuster-model
  • Output Structure
  • Masks
• Minimization Solvers
  • Available Solvers
  • How ADABK Works
  • Termination Control (ADABK solvers)
  • Conditioning
  • Minimizing Programmatically

Analysis

• r_analysis
  • Subcommand Structure
  • Typical Workflow
  • Common Arguments

Tutorials

• FGBuster vs FURAX: Framework Comparison for CMB Component Separation
• Framework Validation: Comparing FURAX and FGBuster
• Advanced Optimization with JAX: Using Optax
• Adaptive Component Separation with K-means Clustering
• Multi-Resolution Component Separation with HEALPix ud_grade
• Visualize Clusters
• Complete Workflow: Scripts and R-statistic Analysis
• Optimizer Benchmark

API Reference

• API Reference
  • Clustering
  • Noise
  • Optimization
  • Binning
  • Data