Reproduce “TorchCurves: Differentiable Parametric Curves” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Alex Shtoff
TorchCurves packages a differentiable PyTorch curve primitive mapping scalar inputs to vector outputs. It evaluates batches using B-spline or Legendre bases, with gradients through inputs and coefficients that may come from another network. In a CPU KAN comparison, the step-time ratio against EfficientKAN was 0.32 at 64 coefficients, width 32, and batch size 1024; timings varied by setting, and architectures differed.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.