QML building blocks - Haiqu

A catalog of the quantum machine learning components shipped with Haiqu SDK. Each entry links to the workflow guide that shows it in context and to the API reference for exact signatures and return types.

Gates and ansatze

Component	Description	Learn more
`su2_equivariant_2_qubit_gate`	The unique 2-qubit SU(2)-equivariant primitive gate: phases the singlet sector and leaves the triplet sector unchanged.	Guide · Reference
`su2_equivariant_3_qubit_gate`	The exact 3-qubit SU(2)-equivariant gate; four angles parameterize the allowed mixing between the two spin-1/2 copies.	Guide · Reference
`su2_equivariant_ansatz`	Parameterized spin-symmetry-preserving ansatz built from the 2-qubit primitive; supports brickwork, linear, and custom pair layouts.	Guide · Reference

Symmetry and verification helpers

Component	Description	Learn more
`is_su2_equivariant`	Check whether a dense unitary commutes with all three global spin generators.	Guide · Reference
`spin_generators`	Dense global spin generators `(S_x, S_y, S_z)` for diagnostics and symmetry-aware observables.	Guide · Reference
`total_spin_ops`	Dense total spin operators `(S^2, S_z)`.	Guide · Reference
`brickwork_pattern`	Nearest-neighbor even-then-odd qubit pair layout used by the brickwork ansatz.	Guide · Reference

Variational solvers

Solver	Description	Learn more
`haiqu.variational_optimization`	Minimize an observable expectation value, or a nonlinear objective over several observables, for a parameterized ansatz using NFT or SciPy optimizers.	Guide · Reference
`haiqu.pretrain`	Pretrain ansatz parameters against the problem objective to warm-start a variational optimization.	Reference
`haiqu.su2_equivariant_compilation`	Fit a shallow brickwork circuit of 2-qubit SU(2)-equivariant gates to an equivariant target unitary.	Guide · Reference