Copyright | (c) Edward Kmett 2010-2015 |
---|---|
License | BSD3 |
Maintainer | ekmett@gmail.com |
Stability | experimental |
Portability | GHC only |
Safe Haskell | None |
Language | Haskell2010 |
Forward mode automatic differentiation
Synopsis
- data Forward a
- auto :: Mode t => Scalar t -> t
- grad :: (Traversable f, Num a) => (f (Forward a) -> Forward a) -> f a -> f a
- grad' :: (Traversable f, Num a) => (f (Forward a) -> Forward a) -> f a -> (a, f a)
- gradWith :: (Traversable f, Num a) => (a -> a -> b) -> (f (Forward a) -> Forward a) -> f a -> f b
- gradWith' :: (Traversable f, Num a) => (a -> a -> b) -> (f (Forward a) -> Forward a) -> f a -> (a, f b)
- jacobian :: (Traversable f, Traversable g, Num a) => (f (Forward a) -> g (Forward a)) -> f a -> g (f a)
- jacobian' :: (Traversable f, Traversable g, Num a) => (f (Forward a) -> g (Forward a)) -> f a -> g (a, f a)
- jacobianWith :: (Traversable f, Traversable g, Num a) => (a -> a -> b) -> (f (Forward a) -> g (Forward a)) -> f a -> g (f b)
- jacobianWith' :: (Traversable f, Traversable g, Num a) => (a -> a -> b) -> (f (Forward a) -> g (Forward a)) -> f a -> g (a, f b)
- jacobianT :: (Traversable f, Functor g, Num a) => (f (Forward a) -> g (Forward a)) -> f a -> f (g a)
- jacobianWithT :: (Traversable f, Functor g, Num a) => (a -> a -> b) -> (f (Forward a) -> g (Forward a)) -> f a -> f (g b)
- hessianProduct :: (Traversable f, Num a) => (f (On (Forward (Forward a))) -> On (Forward (Forward a))) -> f (a, a) -> f a
- hessianProduct' :: (Traversable f, Num a) => (f (On (Forward (Forward a))) -> On (Forward (Forward a))) -> f (a, a) -> f (a, a)
- diff :: Num a => (Forward a -> Forward a) -> a -> a
- diff' :: Num a => (Forward a -> Forward a) -> a -> (a, a)
- diffF :: (Functor f, Num a) => (Forward a -> f (Forward a)) -> a -> f a
- diffF' :: (Functor f, Num a) => (Forward a -> f (Forward a)) -> a -> f (a, a)
- du :: (Functor f, Num a) => (f (Forward a) -> Forward a) -> f (a, a) -> a
- du' :: (Functor f, Num a) => (f (Forward a) -> Forward a) -> f (a, a) -> (a, a)
- duF :: (Functor f, Functor g, Num a) => (f (Forward a) -> g (Forward a)) -> f (a, a) -> g a
- duF' :: (Functor f, Functor g, Num a) => (f (Forward a) -> g (Forward a)) -> f (a, a) -> g (a, a)
Documentation
Forward
mode AD
Instances
Gradient
grad :: (Traversable f, Num a) => (f (Forward a) -> Forward a) -> f a -> f a Source #
Compute the gradient of a function using forward mode AD.
Note, this performs O(n) worse than grad
for n
inputs, in exchange for better space utilization.
grad' :: (Traversable f, Num a) => (f (Forward a) -> Forward a) -> f a -> (a, f a) Source #
Compute the gradient and answer to a function using forward mode AD.
Note, this performs O(n) worse than grad'
for n
inputs, in exchange for better space utilization.
gradWith :: (Traversable f, Num a) => (a -> a -> b) -> (f (Forward a) -> Forward a) -> f a -> f b Source #
Compute the gradient of a function using forward mode AD and combine the result with the input using a user-specified function.
Note, this performs O(n) worse than gradWith
for n
inputs, in exchange for better space utilization.
gradWith' :: (Traversable f, Num a) => (a -> a -> b) -> (f (Forward a) -> Forward a) -> f a -> (a, f b) Source #
Compute the gradient of a function using forward mode AD and the answer, and combine the result with the input using a user-specified function.
Note, this performs O(n) worse than gradWith'
for n
inputs, in exchange for better space utilization.
>>>
gradWith' (,) sum [0..4]
(10,[(0,1),(1,1),(2,1),(3,1),(4,1)])
Jacobian
jacobian :: (Traversable f, Traversable g, Num a) => (f (Forward a) -> g (Forward a)) -> f a -> g (f a) Source #
jacobian' :: (Traversable f, Traversable g, Num a) => (f (Forward a) -> g (Forward a)) -> f a -> g (a, f a) Source #
Compute the Jacobian using Forward
mode AD
along with the actual answer.
jacobianWith :: (Traversable f, Traversable g, Num a) => (a -> a -> b) -> (f (Forward a) -> g (Forward a)) -> f a -> g (f b) Source #
Compute the Jacobian using Forward
mode AD
and combine the output with the input. This must transpose the result, so jacobianWithT
is faster, and allows more result types.
jacobianWith' :: (Traversable f, Traversable g, Num a) => (a -> a -> b) -> (f (Forward a) -> g (Forward a)) -> f a -> g (a, f b) Source #
Compute the Jacobian using Forward
mode AD
combined with the input using a user specified function, along with the actual answer.
Transposed Jacobian
jacobianT :: (Traversable f, Functor g, Num a) => (f (Forward a) -> g (Forward a)) -> f a -> f (g a) Source #
A fast, simple, transposed Jacobian computed with forward-mode AD.
jacobianWithT :: (Traversable f, Functor g, Num a) => (a -> a -> b) -> (f (Forward a) -> g (Forward a)) -> f a -> f (g b) Source #
A fast, simple, transposed Jacobian computed with Forward
mode AD
that combines the output with the input.
Hessian Product
hessianProduct :: (Traversable f, Num a) => (f (On (Forward (Forward a))) -> On (Forward (Forward a))) -> f (a, a) -> f a Source #
Compute the product of a vector with the Hessian using forward-on-forward-mode AD.
hessianProduct' :: (Traversable f, Num a) => (f (On (Forward (Forward a))) -> On (Forward (Forward a))) -> f (a, a) -> f (a, a) Source #
Compute the gradient and hessian product using forward-on-forward-mode AD.
Derivatives
diff :: Num a => (Forward a -> Forward a) -> a -> a Source #
The diff
function calculates the first derivative of a scalar-to-scalar function by forward-mode AD
>>>
diff sin 0
1.0
Directional Derivatives
du :: (Functor f, Num a) => (f (Forward a) -> Forward a) -> f (a, a) -> a Source #
Compute the directional derivative of a function given a zipped up Functor
of the input values and their derivatives
du' :: (Functor f, Num a) => (f (Forward a) -> Forward 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