Phasing via Cost function

from qualtran import Bloq, CompositeBloq, BloqBuilder, Signature, Register
from qualtran import QBit, QInt, QUInt, QAny
from qualtran.drawing import show_bloq, show_call_graph, show_counts_sigma
from typing import *
import numpy as np
import sympy
import cirq

PhasingViaCostFunction

Phases every basis state \(|x\rangle\) by an amount proportional to a cost function \(f(x)\)

This Bloq implements a unitary \(U_f(\gamma)\) which phases each computational state on which the wave-function has support, by an amount proportional to a function of that computational basis state. The general unitary can be defined as

\[ U_f(\gamma) = \sum_{x=0}^{N-1} e^{i 2 \pi \gamma f(x)} |x\rangle \langle x| \]

The strategy to implement \(U_f(\gamma)\) is to use two oracles \(O_\text{direct}\) and \(O_\text{phase}\) s.t.

\[ U_f(\gamma) = O_\text{direct}^\dagger(\mathbb{I}\otimes O_\text{phase})O_\text{direct} \]

\(O^\text{direct}\) evaluates a \(b_\text{direct}\)-bit approximation of the cost function \(f(x)\) and stores it in a new output cost register. Note that the cost register can represent arbitrary fixed point values and be of type QFxp(b_direct, n_frac, signed).

\[ O^\text{direct}|x\rangle|0\rangle^{\otimes b_\text{direct}}_\text{cost}=|x\rangle|f(x)\rangle \]

\(O^\text{phase}\) acts on the cost register computed by \(O^\text{direct}\) and phases the state \(|f(x)\rangle\) by \(e^{i 2\pi \gamma f(x)}\)

\[ O^\text{phase}(\gamma)=\sum_{k=0}^{2^{b_\text{direct}}-1}e^{i 2\pi\gamma k}|k\rangle\langle k| \]

Different strategies for implementing the two oracles would give different costs tradeoffs. See QvrZPow and QvrPhaseGradient for two different implementations of phase oracles described in the reference.

Parameters

• cost_eval_oracle: Cost function evaluation oracle. Must compute the cost in a newly allocated RIGHT register.

• phase_oracle: Oracle to phase the cost register. Must consume the cost register allocated by cost_eval_oracle as a THRU input.

References

• Compilation of Fault-Tolerant Quantum Heuristics for Combinatorial Optimization. Appendix C: Oracles for phasing by cost function

from qualtran.bloqs.rotations.phasing_via_cost_function import PhasingViaCostFunction

Example Instances

from qualtran import QFxp, Register
from qualtran.bloqs.arithmetic import Square
from qualtran.bloqs.rotations.quantum_variable_rotation import QvrZPow

n, gamma, eps = 5, 0.1234, 1e-8
cost_reg = Register('result', QFxp(2 * n, 2 * n, signed=False))
cost_eval_oracle = Square(n)
phase_oracle = QvrZPow(cost_reg, gamma, eps)
square_via_zpow_phasing = PhasingViaCostFunction(cost_eval_oracle, phase_oracle)

from qualtran import QFxp, Register
from qualtran.bloqs.arithmetic import Square
from qualtran.bloqs.rotations.quantum_variable_rotation import QvrPhaseGradient

n, gamma, eps = 5, 0.1234, 1e-8
cost_reg = Register('result', QFxp(2 * n, 2 * n, signed=False))
cost_eval_oracle = Square(n)
phase_oracle = QvrPhaseGradient(cost_reg, gamma, eps)
square_via_phase_gradient = PhasingViaCostFunction(cost_eval_oracle, phase_oracle)

Graphical Signature

from qualtran.drawing import show_bloqs
show_bloqs([square_via_phase_gradient, square_via_zpow_phasing],
           ['`square_via_phase_gradient`', '`square_via_zpow_phasing`'])

Call Graph

from qualtran.resource_counting.generalizers import ignore_split_join
square_via_phase_gradient_g, square_via_phase_gradient_sigma = square_via_phase_gradient.call_graph(max_depth=1, generalizer=ignore_split_join)
show_call_graph(square_via_phase_gradient_g)
show_counts_sigma(square_via_phase_gradient_sigma)

Counts totals:

• QvrPhaseGradient(cost_reg=Register(name='result', dtype=QFxp(bitsize=10, num_frac=10, signed=False), shape=(), side=<Side.THRU: 3>), gamma=0.1234, eps=1e-08): 1

• Square(bitsize=5, uncompute=False): 1

• Square(bitsize=5, uncompute=True): 1

square_via_zpow_phasing_g, square_via_zpow_phasing_sigma = square_via_zpow_phasing.call_graph()
show_call_graph(square_via_zpow_phasing_g)
show_counts_sigma(square_via_zpow_phasing_sigma)

Counts totals:

• TGate(): 160

• Z**0.1234: 10