Numeric.RAD

Reverse Mode Automatic Differentiation via overloading to perform nonstandard interpretation that replaces original numeric type with a bundle that contains a value of the original type and the tape that will be used to recover the value of the sensitivity.

This package uses StableNames internally to recover sharing information from the tape to avoid combinatorial explosion, and thus runs asymptotically faster than it could without such sharing information, but the use of side-effects contained herein is benign.

The API has been built to be close to the design of FAD from the fad package by Barak Pearlmutter and Jeffrey Mark Siskind and contains portions of that code, with minor liberties taken.

First-Order Reverse Mode Automatic Differentiation

data RAD s a

Instances

Bounded a => Bounded (RAD s a)

(Num a, Enum a) => Enum (RAD s a)

Eq a => Eq (RAD s a)

Floating a => Floating (RAD s a)

Fractional a => Fractional (RAD s a)

Num a => Num (RAD s a)

Ord a => Ord (RAD s a)

Real a => Real (RAD s a)

RealFloat a => RealFloat (RAD s a)

RealFrac a => RealFrac (RAD s a)

Show a => Show (RAD s a)

MuRef (RAD s a)

lift :: a -> RAD s a

The lift function injects a primal number into the RAD data type with a 0 derivative. If reverse-mode AD numbers formed a monad, then lift would be return.

First-Order Differentiation Operators

diffUU :: Num a => (forall s. RAD s a -> RAD s a) -> a -> a

The diffUU function calculates the first derivative of a scalar-to-scalar function.

diffUF :: (Functor f, Num a) => (forall s. RAD s a -> f (RAD s a)) -> a -> f a

The diffUF function calculates the first derivative of scalar-to-nonscalar function.

diff2UU :: Num a => (forall s. RAD s a -> RAD s a) -> a -> (a, a)

The diff2UU function calculates the value and derivative, as a pair, of a scalar-to-scalar function.

diff2UF :: (Functor f, Num a) => (forall s. RAD s a -> f (RAD s a)) -> a -> f (a, a)

Note that the signature differs from that used in Numeric.FAD, because while you can always unzip an arbitrary functor, not all functors can be zipped.

Common access patterns

diff :: Num a => (forall s. RAD s a -> RAD s a) -> a -> a

The diff function is a synonym for diffUU.

diff2 :: Num a => (forall s. RAD s a -> RAD s a) -> a -> (a, a)

The diff2 function is a synonym for diff2UU.

jacobian :: (Traversable f, Functor g, Num a) => (forall s. f (RAD s a) -> g (RAD s a)) -> f a -> g (f a)

The jacobian function calcualtes the Jacobian of a nonscalar-to-nonscalar function, using m invocations of reverse AD, where m is the output dimensionality. When the output dimensionality is significantly greater than the input dimensionality you should use jacobian instead.

jacobian2 :: (Traversable f, Functor g, Num a) => (forall s. f (RAD s a) -> g (RAD s a)) -> f a -> g (a, f a)

The jacobian2 function calcualtes both the result and the Jacobian of a nonscalar-to-nonscalar function, using m invocations of reverse AD, where m is the output dimensionality. 'fmap snd' on the result will recover the result of jacobian

grad :: (Traversable f, Num a) => (forall s. f (RAD s a) -> RAD s a) -> f a -> f a

grad2 :: (Traversable f, Num a) => (forall s. f (RAD s a) -> RAD s a) -> f a -> (a, f a)

Optimization Routines

zeroNewton :: Fractional a => (forall s. RAD s a -> RAD s a) -> a -> [a]

The zeroNewton function finds a zero of a scalar function using Newton's method; its output is a stream of increasingly accurate results. (Modulo the usual caveats.)

TEST CASE: take 10 $ zeroNewton (\x->x^2-4) 1 -- converge to 2.0

TEST CASE :module Data.Complex Numeric.RAD take 10 $ zeroNewton ((+1).(^2)) (1 :+ 1) -- converge to (0 :+ 1)

inverseNewton :: Fractional a => (forall s. RAD s a -> RAD s a) -> a -> a -> [a]

The inverseNewton function inverts a scalar function using Newton's method; its output is a stream of increasingly accurate results. (Modulo the usual caveats.)

TEST CASE: take 10 $ inverseNewton sqrt 1 (sqrt 10) -- converge to 10

fixedPointNewton :: Fractional a => (forall s. RAD s a -> RAD s a) -> a -> [a]

The fixedPointNewton function find a fixedpoint of a scalar function using Newton's method; its output is a stream of increasingly accurate results. (Modulo the usual caveats.)

extremumNewton :: Fractional a => (forall s t. RAD t (RAD s a) -> RAD t (RAD s a)) -> a -> [a]

The extremumNewton function finds an extremum of a scalar function using Newton's method; produces a stream of increasingly accurate results. (Modulo the usual caveats.)

argminNaiveGradient :: (Fractional a, Ord a) => (forall s. [RAD s a] -> RAD s a) -> [a] -> [[a]]

The argminNaiveGradient function performs a multivariate optimization, based on the naive-gradient-descent in the file stalingrad/examples/flow-tests/pre-saddle-1a.vlad from the VLAD compiler Stalingrad sources. Its output is a stream of increasingly accurate results. (Modulo the usual caveats.) This is O(n) faster than argminNaiveGradient