dfsc

dfsc is a research-grade beta for differentiable Mittag-Leffler propagation in PyTorch. It combines trainable fractional orders, dense and matrix-function operator paths, reliability diagnostics, history-aware fallbacks, and neural residual composition under one namespace.

The validated strength of the package is known- or approximated-propagator workloads. It is not a universal fractional differential-equation solver.

Start here

• Public API
• Application domains and assumptions
• Real-data evidence
• Current maturity
• Release checklist

Installation, quick-use examples, and the complete component map are maintained in the repository README.md and packaged distributions.