Portability  GHC only 

Stability  experimental 
Maintainer  ekmett@gmail.com 
Safe Haskell  None 
 Gradients (Reverse Mode)
 Higher Order Gradients (SparseonReverse)
 Jacobians (Sparse or Reverse)
 Higher Order Jacobian (SparseonReverse)
 Transposed Jacobians (Forward Mode)
 Hessian (SparseOnReverse)
 Hessian Tensors (Sparse or SparseOnReverse)
 Hessian Tensors (Sparse)
 Hessian Vector Products (ForwardOnReverse)
 Derivatives (Forward Mode)
 Derivatives (Tower)
 Directional Derivatives (Forward Mode)
 Directional Derivatives (Tower)
 Taylor Series (Tower)
 Maclaurin Series (Tower)
 Gradient Descent
MixedMode Automatic Differentiation.
Each combinator exported from this module chooses an appropriate AD mode. The following basic operations are supported, modified as appropriate by the suffixes below:

grad
computes the gradient (partial derivatives) of a function at a point 
jacobian
computes the Jacobian matrix of a function at a point 
diff
computes the derivative of a function at a point 
du
computes a directional derivative of a function at a point 
hessian
compute the Hessian matrix (matrix of second partial derivatives) of a function at a point
The suffixes have the following meanings:

'
 also return the answer 
With
lets the user supply a function to blend the input with the output 
F
is a version of the base function lifted to return aTraversable
(orFunctor
) result 
s
means the function returns all higher derivatives in a list or fbranchingStream

T
means the result is transposed with respect to the traditional formulation. 
0
means that the resulting derivative list is padded with 0s at the end.
 grad :: (Traversable f, Num a) => (forall s. Mode s => f (AD s a) > AD s a) > f a > f a
 grad' :: (Traversable f, Num a) => (forall s. Mode s => f (AD s a) > AD s a) > f a > (a, f a)
 gradWith :: (Traversable f, Num a) => (a > a > b) > (forall s. Mode s => f (AD s a) > AD s a) > f a > f b
 gradWith' :: (Traversable f, Num a) => (a > a > b) > (forall s. Mode s => f (AD s a) > AD s a) > f a > (a, f b)
 grads :: (Traversable f, Num a) => (forall s. Mode s => f (AD s a) > AD s a) > f a > Cofree f a
 jacobian :: (Traversable f, Functor g, Num a) => (forall s. Mode s => f (AD s a) > g (AD s a)) > f a > g (f a)
 jacobian' :: (Traversable f, Functor g, Num a) => (forall s. Mode s => f (AD s a) > g (AD s a)) > f a > g (a, f a)
 jacobianWith :: (Traversable f, Functor g, Num a) => (a > a > b) > (forall s. Mode s => f (AD s a) > g (AD s a)) > f a > g (f b)
 jacobianWith' :: (Traversable f, Functor g, Num a) => (a > a > b) > (forall s. Mode s => f (AD s a) > g (AD s a)) > f a > g (a, f b)
 jacobians :: (Traversable f, Functor g, Num a) => (forall s. Mode s => f (AD s a) > g (AD s a)) > f a > g (Cofree f a)
 jacobianT :: (Traversable f, Functor g, Num a) => (forall s. Mode s => f (AD s a) > g (AD s a)) > f a > f (g a)
 jacobianWithT :: (Traversable f, Functor g, Num a) => (a > a > b) > (forall s. Mode s => f (AD s a) > g (AD s a)) > f a > f (g b)
 hessian :: (Traversable f, Num a) => (forall s. Mode s => f (AD s a) > AD s a) > f a > f (f a)
 hessian' :: (Traversable f, Num a) => (forall s. Mode s => f (AD s a) > AD s a) > f a > (a, f (a, f a))
 hessianF :: (Traversable f, Functor g, Num a) => (forall s. Mode s => f (AD s a) > g (AD s a)) > f a > g (f (f a))
 hessianF' :: (Traversable f, Functor g, Num a) => (forall s. Mode s => f (AD s a) > g (AD s a)) > f a > g (a, f (a, f a))
 hessianProduct :: (Traversable f, Num a) => (forall s. Mode s => f (AD s a) > AD s a) > f (a, a) > f a
 hessianProduct' :: (Traversable f, Num a) => (forall s. Mode s => f (AD s a) > AD s a) > f (a, a) > f (a, a)
 diff :: Num a => (forall s. Mode s => AD s a > AD s a) > a > a
 diffF :: (Functor f, Num a) => (forall s. Mode s => AD s a > f (AD s a)) > a > f a
 diff' :: Num a => (forall s. Mode s => AD s a > AD s a) > a > (a, a)
 diffF' :: (Functor f, Num a) => (forall s. Mode s => AD s a > f (AD s a)) > a > f (a, a)
 diffs :: Num a => (forall s. Mode s => AD s a > AD s a) > a > [a]
 diffsF :: (Functor f, Num a) => (forall s. Mode s => AD s a > f (AD s a)) > a > f [a]
 diffs0 :: Num a => (forall s. Mode s => AD s a > AD s a) > a > [a]
 diffs0F :: (Functor f, Num a) => (forall s. Mode s => AD s a > f (AD s a)) > a > f [a]
 du :: (Functor f, Num a) => (forall s. Mode s => f (AD s a) > AD s a) > f (a, a) > a
 du' :: (Functor f, Num a) => (forall s. Mode s => f (AD s a) > AD s a) > f (a, a) > (a, a)
 duF :: (Functor f, Functor g, Num a) => (forall s. Mode s => f (AD s a) > g (AD s a)) > f (a, a) > g a
 duF' :: (Functor f, Functor g, Num a) => (forall s. Mode s => f (AD s a) > g (AD s a)) > f (a, a) > g (a, a)
 dus :: (Functor f, Num a) => (forall s. Mode s => f (AD s a) > AD s a) > f [a] > [a]
 dus0 :: (Functor f, Num a) => (forall s. Mode s => f (AD s a) > AD s a) > f [a] > [a]
 dusF :: (Functor f, Functor g, Num a) => (forall s. Mode s => f (AD s a) > g (AD s a)) > f [a] > g [a]
 dus0F :: (Functor f, Functor g, Num a) => (forall s. Mode s => f (AD s a) > g (AD s a)) > f [a] > g [a]
 taylor :: Fractional a => (forall s. Mode s => AD s a > AD s a) > a > a > [a]
 taylor0 :: Fractional a => (forall s. Mode s => AD s a > AD s a) > a > a > [a]
 maclaurin :: Fractional a => (forall s. Mode s => AD s a > AD s a) > a > [a]
 maclaurin0 :: Fractional a => (forall s. Mode s => AD s a > AD s a) > a > [a]
 gradientDescent :: (Traversable f, Fractional a, Ord a) => (forall s. Mode s => f (AD s a) > AD s a) > f a > [f a]
 gradientAscent :: (Traversable f, Fractional a, Ord a) => (forall s. Mode s => f (AD s a) > AD s a) > f a > [f a]
 conjugateGradientDescent :: (Traversable f, Fractional a, Ord a) => (forall s. Mode s => f (AD s a) > AD s a) > f a > [f a]
 conjugateGradientAscent :: (Traversable f, Fractional a, Ord a) => (forall s. Mode s => f (AD s a) > AD s a) > f a > [f a]
Gradients (Reverse Mode)
grad :: (Traversable f, Num a) => (forall s. Mode s => f (AD s a) > AD s a) > f a > f aSource
The grad
function calculates the gradient of a nonscalartoscalar function with reversemode AD in a single pass.
>>>
grad (\[x,y,z] > x*y+z) [1,2,3]
[2,1,1]
grad' :: (Traversable f, Num a) => (forall s. Mode s => f (AD s a) > AD s a) > f a > (a, f a)Source
The grad'
function calculates the result and gradient of a nonscalartoscalar function with reversemode AD in a single pass.
>>>
grad' (\[x,y,z] > x*y+z) [1,2,3]
(5,[2,1,1])
gradWith :: (Traversable f, Num a) => (a > a > b) > (forall s. Mode s => f (AD s a) > AD s a) > f a > f bSource
gradWith' :: (Traversable f, Num a) => (a > a > b) > (forall s. Mode s => f (AD s a) > AD s a) > f a > (a, f b)Source
Higher Order Gradients (SparseonReverse)
grads :: (Traversable f, Num a) => (forall s. Mode s => f (AD s a) > AD s a) > f a > Cofree f aSource
Jacobians (Sparse or Reverse)
jacobian :: (Traversable f, Functor g, Num a) => (forall s. Mode s => f (AD s a) > g (AD s a)) > f a > g (f a)Source
jacobian' :: (Traversable f, Functor g, Num a) => (forall s. Mode s => f (AD s a) > g (AD s a)) > f a > g (a, f a)Source
jacobianWith :: (Traversable f, Functor g, Num a) => (a > a > b) > (forall s. Mode s => f (AD s a) > g (AD s a)) > f a > g (f b)Source
calculates the Jacobian of a nonscalartononscalar function, automatically choosing between forward and reverse mode AD based on the number of inputs and outputs.
jacobianWith
g f
The resulting Jacobian matrix is then recombined elementwise with the input using g
.
If you know the relative number of inputs and outputs, consider jacobianWith
or jacobianWith
.
jacobianWith' :: (Traversable f, Functor g, Num a) => (a > a > b) > (forall s. Mode s => f (AD s a) > g (AD s a)) > f a > g (a, f b)Source
calculates the answer and Jacobian of a nonscalartononscalar function, automatically choosing between sparse and reverse mode AD based on the number of inputs and outputs.
jacobianWith'
g f
The resulting Jacobian matrix is then recombined elementwise with the input using g
.
If you know the relative number of inputs and outputs, consider jacobianWith'
or jacobianWith'
.
Higher Order Jacobian (SparseonReverse)
jacobians :: (Traversable f, Functor g, Num a) => (forall s. Mode s => f (AD s a) > g (AD s a)) > f a > g (Cofree f a)Source
Transposed Jacobians (Forward Mode)
jacobianT :: (Traversable f, Functor g, Num a) => (forall s. Mode s => f (AD s a) > g (AD s a)) > f a > f (g a)Source
A fast, simple, transposed Jacobian computed with forwardmode AD.
jacobianWithT :: (Traversable f, Functor g, Num a) => (a > a > b) > (forall s. Mode s => f (AD s a) > g (AD s a)) > f a > f (g b)Source
Hessian (SparseOnReverse)
hessian :: (Traversable f, Num a) => (forall s. Mode s => f (AD s a) > AD s a) > f a > f (f a)Source
Compute the Hessian via the Jacobian of the gradient. gradient is computed in reverse mode and then the Jacobian is computed in sparse (forward) mode.
hessian' :: (Traversable f, Num a) => (forall s. Mode s => f (AD s a) > AD s a) > f a > (a, f (a, f a))Source
Hessian Tensors (Sparse or SparseOnReverse)
hessianF :: (Traversable f, Functor g, Num a) => (forall s. Mode s => f (AD s a) > g (AD s a)) > f a > g (f (f a))Source
Compute the order 3 Hessian tensor on a nonscalartononscalar function using Sparse
or 'Sparse'on'Reverse'
Hessian Tensors (Sparse)
hessianF' :: (Traversable f, Functor g, Num a) => (forall s. Mode s => f (AD s a) > g (AD s a)) > f a > g (a, f (a, f a))Source
Hessian Vector Products (ForwardOnReverse)
hessianProduct :: (Traversable f, Num a) => (forall s. Mode s => f (AD s a) > AD s a) > f (a, a) > f aSource
computes the product of the hessian hessianProduct
f wvH
of a nonscalartoscalar function f
at w =
with a vector fst
$ wvv = snd $ wv
using "Pearlmutter's method" from http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.29.6143, which states:
H v = (d/dr) grad_w (w + r v)  r = 0
Or in other words, we take the directional derivative of the gradient. The gradient is calculated in reverse mode, then the directional derivative is calculated in forward mode.
hessianProduct' :: (Traversable f, Num a) => (forall s. Mode s => f (AD s a) > AD s a) > f (a, a) > f (a, a)Source
computes both the gradient of a nonscalartoscalar hessianProduct'
f wvf
at w =
and the product of the hessian fst
$ wvH
at w
with a vector v = snd $ wv
using "Pearlmutter's method". The outputs are returned wrapped in the same functor.
H v = (d/dr) grad_w (w + r v)  r = 0
Or in other words, we return the gradient and the directional derivative of the gradient. The gradient is calculated in reverse mode, then the directional derivative is calculated in forward mode.
Derivatives (Forward Mode)
Derivatives (Tower)
Directional Derivatives (Forward Mode)
du :: (Functor f, Num a) => (forall s. Mode s => f (AD s a) > AD s a) > f (a, a) > aSource
Compute the directional derivative of a function given a zipped up Functor
of the input values and their derivatives
du' :: (Functor f, Num a) => (forall s. Mode s => f (AD s a) > AD s a) > f (a, a) > (a, a)Source
Compute the answer and directional derivative of a function given a zipped up Functor
of the input values and their derivatives
duF :: (Functor f, Functor g, Num a) => (forall s. Mode s => f (AD s a) > g (AD s a)) > f (a, a) > g aSource
Compute a vector of directional derivatives for a function given a zipped up Functor
of the input values and their derivatives.
duF' :: (Functor f, Functor g, Num a) => (forall s. Mode s => f (AD s a) > g (AD s a)) > f (a, a) > g (a, a)Source
Compute a vector of answers and directional derivatives for a function given a zipped up Functor
of the input values and their derivatives.
Directional Derivatives (Tower)
dusF :: (Functor f, Functor g, Num a) => (forall s. Mode s => f (AD s a) > g (AD s a)) > f [a] > g [a]Source
dus0F :: (Functor f, Functor g, Num a) => (forall s. Mode s => f (AD s a) > g (AD s a)) > f [a] > g [a]Source
Taylor Series (Tower)
Maclaurin Series (Tower)
maclaurin0 :: Fractional a => (forall s. Mode s => AD s a > AD s a) > a > [a]Source
Gradient Descent
gradientDescent :: (Traversable f, Fractional a, Ord a) => (forall s. Mode s => f (AD s a) > AD s a) > f a > [f a]Source
The gradientDescent
function performs a multivariate
optimization, based on the naivegradientdescent in the file
stalingrad/examples/flowtests/presaddle1a.vlad
from the
VLAD compiler Stalingrad sources. Its output is a stream of
increasingly accurate results. (Modulo the usual caveats.)
It uses reverse mode automatic differentiation to compute the gradient.
gradientAscent :: (Traversable f, Fractional a, Ord a) => (forall s. Mode s => f (AD s a) > AD s a) > f a > [f a]Source
Perform a gradient descent using reverse mode automatic differentiation to compute the gradient.
conjugateGradientDescent :: (Traversable f, Fractional a, Ord a) => (forall s. Mode s => f (AD s a) > AD s a) > f a > [f a]Source
Perform a conjugate gradient descent using reverse mode automatic differentiation to compute the gradient.
conjugateGradientAscent :: (Traversable f, Fractional a, Ord a) => (forall s. Mode s => f (AD s a) > AD s a) > f a > [f a]Source
Perform a conjugate gradient ascent using reverse mode automatic differentiation to compute the gradient.