apply_param_jac: Adjoint of the parameter -> (c, d, A, b) tensor map

In CVXR: Disciplined Convex Optimization

Adjoint of the parameter -> (c, d, A, b) tensor map

Description

Given derivatives ⁠(delc, delA, delb)⁠ of a downstream objective with respect to the conic problem data, returns derivatives with respect to each Parameter (keyed by parameter id). Mirrors ParamConeProg.apply_param_jac in cone_matrix_stuffing.py:242-280.

Usage

apply_param_jac(param_prog, delc, delA, delb, active_params = NULL)

Arguments

param_prog	A ParamConeProg.
delc	Numeric vector of length x_length.
delA	Sparse matrix of shape ⁠(m, x_length)⁠ (same shape as the conic constraint matrix A).
delb	Numeric vector of length m.
active_params	Optional character vector of parameter ids to restrict the output to. Default: all parameters.

Details

Reusing the tensor identity that apply_parameters exploits in the forward direction: c = c_tensor[1:x_length, ] %*% p (omitting the offset row), ⁠vec(A) // b = A_tensor %*% p⁠ (column-major; b is last block), the adjoint is just the transpose of the same tensors applied to the stacked (delc, vec(delA), delb).

Value

A named list mapping as.character(param_id) to a numeric array of the parameter's shape.

CVXR documentation built on Aug. 24, 2026, 9:10 a.m.