4 packages tagged with “autodiff”
A library for parallelized reverse mode automatic differentiation in C# for custom neural network development.
A lightweight library for parallelized reverse mode automatic differentiation in C# for custom neural network development using single-precision.
GPU-accelerated algorithms for DotCompute. Includes FFT, AutoDiff, sparse matrix operations, signal processing, and cryptographic primitives.
A high-performance adjoint algorithmic (automatic) differentiation library designed for computing first- and second-order derivatives (i.e. gradients and Hessians), ideal for numerical optimization and Monte-Carlo applications, such as derivatives pricing.