auto-grad

Explore how Usul turns complex neural networks into simple math

If you're tired of treating neural networks like black boxes summoned by opaque frameworks, Usul-cpp offers a refreshing alternative. This lightweight C++ library lets you express models as pure math, tensors, forward passes, and backpropagation written out in clear, readable code. The XOR example shown here is elegantly simple: define your weights, write the arithmetic, and let Usul handle gradient descent. It's a human-centered approach that demystifies deep learning without sacrificing precision or performance.