alexshtf/torchcurves

Based on the README, here's a technical summary: Implements vectorized parametric curve evaluation (B-splines, Legendre polynomials) with learnable coefficients through custom autograd functions and efficient numerics like Clenshaw recursion and Cox-DeBoor algorithms. Enables shape-constrained modeling by leveraging mathematical properties of curves—monotonicity through ordered control points, bounded outputs via compact interval mapping—without explicit constraints. Designed as composable PyTorch modules that integrate seamlessly into standard nn.Sequential architectures for KAN layers, continuous embeddings, and auction models.

Available on PyPI.