fdm.multivariate module¶

fdm.multivariate.default_adaptive_method = <fdm.fdm.FDM object>¶

Default adaptive method.

fdm.multivariate.default_robust_method = <fdm.fdm.FDM object>¶

Default robust method.

fdm.multivariate.gradient(f, method=<fdm.fdm.FDM object>)[source]¶

Compute the gradient of f.

The function f should be scalar valued.

Parameters

f (function) – Function to compute gradient of. This should be a scalar-valued function.
method (fdm.FDM, optional) – Finite difference method to use. Defaults to multivariate.default_adaptive_method.

Returns

Gradient.

Return type

function

fdm.multivariate.hvp(f, v, jac_method=<fdm.fdm.FDM object>, dir_method=<fdm.fdm.FDM object>)[source]¶

Compute a Hessian-vector product.

Parameters

f (function) – Function to compute Hessian of.
v (tensor) – Vector to multiply Hessian with. Should be of the same form as the arguments supplied to f.
jac_method (fdm.FDM, optional) – Finite difference method to use for Jacobian computation. Defaults to multivariate.default_adaptive_method.
dir_method (fdm.FDM, optional) – Finite difference method to use for directional derivative computation. Defaults to multivariate.default_robust_method.

Returns

Hessian of f multiplied by v.

Return type

function

fdm.multivariate.jacobian(f, method=<fdm.fdm.FDM object>)[source]¶

Compute the Jacobian of f.

Parameters

f (function) – Function to compute Jacobian of.
method (fdm.FDM, optional) – Finite difference method to use. Defaults to multivariate.default_adaptive_method.

Returns

Jacobian.

Return type

function

fdm.multivariate.jvp(f, v, method=<fdm.fdm.FDM object>)[source]¶

Compute a Jacobian-vector product, also known as a directional derivative.

Parameters

f (function) – Function to compute Jacobian of.
v (tensor) – Vector to multiply Jacobian with. Should be of the same form as the arguments supplied to f.
method (fdm.FDM, optional) – Finite difference method to use. Defaults to multivariate.default_adaptive_method.

Returns

Jacobian of f multiplied by v, or directional: derivative in the direction v.

Return type

function