approx_gradient: Approximate Gradient

In cgraph: Computational Graphs

Description

Differentiate a target node with respect to a given node by numerical differentiation.

Usage

approx_gradient(graph, target, node, index = 1, eps = 1e-04)

Arguments

graph	cg_graph object, graph that is differentiated.
target	cg_node object, node in the graph that is differentiated. Alternatively, argument target can be a character scalar denoting the name of the node in the graph that is differentiated.
node	cg_node object, node with respect to which the target node is differentiated. Alternatively, argument node can be a character scalar denoting the name of the node in the graph to which the target node is differentiated.
index	numerical scalar, index of the target node that is differentiated. Defaults to the first element.
eps	numerical scalar, step size. Defaults to 1e-4.

Value

numerical vector or array.

Note

All nodes required to compute the target node must have a value or their value must be able to be computed at run-time. Only those nodes needed to compute the target node (including the target itself) are evaluated.

The graph is differentiation by the symmetric difference quotient. This method can only be used to differentiate scalars. In case the target node evaluates to a vector or an array, argument index can be used to specify which element of the vector or array is differentiated. The derivative has the same shape as the value of node supplied to argument node.

Numerical differentiation is subject to estimation error and can be very slow. Therefore, this function should only be used for testing purposes.

If the name of the node is supplied to argument target or argument node, a linear search is performed to retrieve the nodes from the graph. In case multiple nodes share the same name, the last node added to the graph is retrieved. Please note that this linear search can become relatively expensive for large graphs.

Author(s)

Ron Triepels

cgraph documentation built on Feb. 9, 2020, 5:07 p.m.