Linear operators

Owner import path¶

jaxstro.numerics.operators

Purpose¶

Small PyTree linear operators and explicit algebraic composition.

Public records and callables¶

LinearOperator, DenseOperator, DiagonalOperator, ScaledOperator, SumOperator, ProductOperator, TransposeOperator, BlockDiagonalOperator, scale, add, compose, transpose, and block_diag.

Shape and dtype expectations¶

Operators declare matrix shapes; matvec and rmatvec require compatible trailing vector axes. Component dtypes determine result dtype through JAX.

JAX transforms and AD classification¶

Operator PyTrees compose with JIT, VMAP, and AD when their leaves and shapes are fixed. Composition order is explicit.

Failure behavior¶

Incompatible shapes raise during construction or evaluation. The module does not infer sparsity, conditioning, or solver policy.

Contract and evidence links¶

See Linear operators and Validation.

Canonical import example¶

from jaxstro.numerics.operators import DenseOperator