
Arbitrary-order exact derivatives at machine precision
nabla provides a single composable operator
D that differentiates any R function to any order —
exactly, at machine precision, through loops, branches, and all control
flow:
library(nabla)
f <- function(x) x[1]^2 * exp(x[2])
D(f, c(1, 0)) # gradient
D(f, c(1, 0), order = 2) # Hessian
D(f, c(1, 0), order = 3) # 2×2×2 third-order tensor
D(f, c(1, 0), order = 4) # 2×2×2×2 fourth-order tensorEach application of D adds one dimension to the output.
D(D(f)) gives the Hessian, D(D(D(f))) gives
the third-order tensor, and so on — no limit on order, no loss of
precision, no symbolic algebra.
| Finite Differences | Symbolic Diff | AD (nabla) | |
|---|---|---|---|
| Accuracy | O(h) or O(h²) truncation error | Exact | Exact (machine precision) |
| Higher-order | Error compounds rapidly | Expression swell | Composes cleanly to any order |
| Control flow | Works | Breaks on if/for/while | Works through any code |
Finite differences lose precision at higher orders (each order
multiplies the error). Symbolic differentiation suffers from expression
swell. nabla composes D via nested dual
numbers — each order is as precise as the first.
# Install from CRAN
install.packages("nabla")
# Or install development version from GitHub
remotes::install_github("queelius/nabla")D operatorD is the core of nabla. It differentiates
any function f and returns a new function — which can
itself be differentiated:
f <- function(x) x[1]^2 * x[2] + sin(x[2])
Df <- D(f) # first derivative (function)
DDf <- D(Df) # second derivative (function)
DDDf <- D(DDf) # third derivative (function)
Df(c(3, 4)) # gradient vector
DDf(c(3, 4)) # Hessian matrix
DDDf(c(3, 4)) # 2×2×2 tensorEquivalently, evaluate directly at a point:
D(f, c(3, 4)) # gradient
D(f, c(3, 4), order = 2) # Hessian
D(f, c(3, 4), order = 3) # third-order tensorgradient(), hessian(), and
jacobian() are convenience wrappers:
gradient(f, c(3, 4)) # == D(f, c(3, 4))
hessian(f, c(3, 4)) # == D(f, c(3, 4), order = 2)A dual number extends the reals with an infinitesimal ε where ε² = 0:
\[f(x + \varepsilon) = f(x) + f'(x)\,\varepsilon\]
For higher orders, nabla nests dual numbers: a dual
whose components are themselves duals. Each level of nesting extracts
one additional order of derivative — so D(D(D(f)))
propagates through triply-nested duals to produce exact third
derivatives. This works through lgamma,
psigamma, trig functions, and all of R’s math — no special
cases needed.
optim() and nlminb()D operator, curvature, Taylor
expansiongradient() and
hessian() with optim() and
nlminb()MIT