xad-rs 0.8.2

Exact automatic differentiation for Rust — forward-mode, reverse-mode, first- and second-order, with named variable support and a unified `Real` trait for mode-agnostic numerical code
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

//! Composite AD operations: Jacobian and Hessian computations.
//!
//! Higher-level wrappers that orchestrate the modes in [`crate::forward`]
//! and [`crate::reverse`] to produce the canonical derivative objects
//! (`J ∈ R^{m × n}`, `H ∈ R^{n × n}`) without users wiring tapes or
//! seed matrices by hand. `compute_full_hessian` is exact (one
//! `Jet2Vec` forward pass); `compute_hessian` is the finite-difference
//! variant for when `Jet2Vec` is not an option.
//!
//! See also (theory): [`docs/theory/04-second-order-and-k-jets.md`](https://github.com/sercanatalik/xad-rs/blob/main/docs/theory/04-second-order-and-k-jets.md).

pub mod jacobian;
pub mod hessian;

pub use jacobian::{compute_jacobian_rev, compute_jacobian_fwd, NamedJacobian, compute_named_jacobian};
pub use hessian::{compute_hessian, NamedHessian, compute_full_hessian};