Complex Tensor Handler

About

This skill provides specialized patterns and best practices for working with complex-valued data in PyTorch, specifically tailored for astronomical imaging applications. It streamlines tasks such as 2D Fourier transforms (FFT/IFFT), conversions between phase/amplitude and real/imaginary representations, and the implementation of complex-valued loss functions. By addressing common pitfalls like phase wrapping and data type mismatches, it ensures robust mathematical operations in frequency-domain analysis and complex-valued neural network architectures.

Key Features

• Complex-aware loss function implementations including MSE and phase-amplitude loss
• Conversion patterns between phase/amplitude and real/imaginary representations
• 0 GitHub stars
• Visualization utilities for complex images using HSV phase and grayscale amplitude maps
• Implementation of 2D FFT and IFFT operations with proper normalization and shifting
• Best practices for avoiding common pitfalls like phase wrapping and non-complex-safe activations

Use Cases

• Processing astronomical measurements that involve both phase and amplitude data
• Implementing frequency-domain filtering and signal processing pipelines in PyTorch
• Developing and training complex-valued neural networks for image reconstruction