Draft · v0.2 · Unreviewed
← Index/Project · Nadarasa ReductionDraft · v0.2 · Selene Emulator

primitive · G10

Quantum Signal Processing — execution layer

Single signal qubit. U(x) = e0Z · Πk[ W(x) · ekZ ] with W(x) = ei·acos(x)·X. Hand-chosen phase sequence φ = (π/3, π/3, π/3, π/3) — d = 3, no phase finder. The host computes P(0) = |⟨0|U|0⟩|² exactly via NumPy 2×2 multiplication; Selene gives the empirical rate.

qsp_verified · worst |Δ| = 0.0176 · 2048 shots × 8 x-points

← frontier card

Measured vs reference P(measure = 0)

Per-x table

xmeasured P(0)reference P(0)Δ
-1.0001.00001.00000.0000
-0.5000.56050.5781-0.0176
-0.2500.16650.1760-0.0095
0.0000.00000.0000-0.0000
0.2500.16700.1760-0.0090
0.5000.57370.5781-0.0044
0.7500.92040.91630.0042
1.0001.00001.00000.0000

Kernel snippet (x = 0.5)

open in Playpond ↗

from quantum.nadarasa_g10_lib import (
    guppy, qubit, h, rx, rz, measure, result, angle,
)

@guppy
def program() -> None:
    s = qubit()
    rz(s, angle(-0.6666666666666666))
    rx(s, angle(-0.6666666666666667))
    rz(s, angle(-0.6666666666666666))
    rx(s, angle(-0.6666666666666667))
    rz(s, angle(-0.6666666666666666))
    rx(s, angle(-0.6666666666666667))
    rz(s, angle(-0.6666666666666666))
    result("s", measure(s))
Single-signal-qubit QSP with a HAND-CHOSEN phase sequence (d = 3, φ = (π/3, π/3, π/3, π/3)). U(x) = e^{iφ_0 Z} · Π_k [ W(x) · e^{iφ_k Z} ] with W(x) = e^{i·acos(x)·X}. The host computes P(measure = 0) = |⟨0| U |0⟩|² in closed form via NumPy 2×2 matrix multiplication and compares to the empirical Selene rate at each x. Verdict = verified if worst-case |measured − reference| stays below 0.04. The card stays at verified for the EXECUTION layer; the phase-finder layer (target polynomial → φ via convex optimization) remains the open frontier work.