PRAssisted

Role: Stateful two-party PR-assisted resource that returns a uniformly random first outcome string and a second outcome string biased-correlated to the first based on both parties’ measurement settings and p_high.

Location: Q_Sea_Battle.pr_assisted.PRAssisted

Constructor

Parameter	Type	Description
length	int, constraint \(length \ge 1\)	Number of bits in each measurement/outcome string.
p_high	float, constraint \(0.0 \le p\_high \le 1.0\)	Correlation parameter controlling how likely the second outcome matches (or complements) the first outcome per index.

Preconditions

• length is int, constraint \(length \ge 1\).
• p_high is int or float, convertible to float, constraint \(0.0 \le p\_high \le 1.0\).

Postconditions

• self.length is set to length.
• self.p_high is set to float(p_high).
• Measurement state is reset: a_measured == False, b_measured == False, and previous-measurement/outcome fields are None.
• A NumPy RNG is created and stored in self._rng.

Errors

• Raises TypeError if length is not an int.
• Raises ValueError if length < 1.
• Raises TypeError if p_high is not an int or float.
• Raises ValueError if p_high is outside \([0.0, 1.0]\).

Example

Construct and use in a single round

import numpy as np
from Q_Sea_Battle.pr_assisted import PRAssisted

pr = PRAssisted(length=4, p_high=0.8)

x_a = np.array([0, 1, 0, 1], dtype=int)
x_b = np.array([1, 1, 0, 0], dtype=int)

out_a = pr.measurement_a(x_a)
out_b = pr.measurement_b(x_b)

pr.reset()

Public Methods

measurement_a(measurement)

Perform a measurement by party A.

Arguments

• measurement: np.ndarray, dtype int, values in {0,1}, shape (length,), 1D measurement setting vector.

Returns

• np.ndarray, dtype int, values in {0,1}, shape (length,), 1D outcome vector for party A.

Preconditions

• self.a_measured is False for the current round.
• measurement is valid per _validate_measurement: coercible to 1D int array, shape (length,), values in {0,1}.

Postconditions

• self.a_measured is set to True.
• If this is the first measurement of the round: prev_party, prev_measurement, and prev_outcome are stored and the returned outcome is uniformly random in {0,1}^{length}.
• If this is the second measurement of the round: the returned outcome is generated by _second_measurement(...) using stored previous measurement/outcome.

Errors

• Raises ValueError if A already measured in the current round.
• Raises ValueError if measurement is not 1D, not shape (length,), or contains values outside {0,1}.

Example

A measures first

import numpy as np
from Q_Sea_Battle.pr_assisted import PRAssisted

pr = PRAssisted(length=3, p_high=0.9)
out_a = pr.measurement_a(np.array([0, 0, 1], dtype=int))

measurement_b(measurement)

Perform a measurement by party B.

Arguments

• measurement: np.ndarray, dtype int, values in {0,1}, shape (length,), 1D measurement setting vector.

Returns

• np.ndarray, dtype int, values in {0,1}, shape (length,), 1D outcome vector for party B.

Preconditions

• self.b_measured is False for the current round.
• measurement is valid per _validate_measurement: coercible to 1D int array, shape (length,), values in {0,1}.

Postconditions

• self.b_measured is set to True.
• If this is the first measurement of the round: prev_party, prev_measurement, and prev_outcome are stored and the returned outcome is uniformly random in {0,1}^{length}.
• If this is the second measurement of the round: the returned outcome is generated by _second_measurement(...) using stored previous measurement/outcome.

Errors

• Raises ValueError if B already measured in the current round.
• Raises ValueError if measurement is not 1D, not shape (length,), or contains values outside {0,1}.

Example

B measures second

import numpy as np
from Q_Sea_Battle.pr_assisted import PRAssisted

pr = PRAssisted(length=2, p_high=0.7)
out_a = pr.measurement_a(np.array([1, 0], dtype=int))
out_b = pr.measurement_b(np.array([1, 1], dtype=int))

reset()

Reset internal measurement state for reuse of the resource.

Arguments

• None.

Returns

• None.

Postconditions

• a_measured == False
• b_measured == False
• prev_party is None
• prev_measurement is None
• prev_outcome is None

Errors

• Not specified.

Example

Reset between rounds

from Q_Sea_Battle.pr_assisted import PRAssisted

pr = PRAssisted(length=1, p_high=0.5)
pr.reset()

Data & State

Public attributes

• length: int, constraint \(length \ge 1\), scalar, number of bits in each measurement/outcome string.
• p_high: float, constraint \(0.0 \le p\_high \le 1.0\), scalar, correlation parameter.
• a_measured: bool, scalar, whether party A has measured in the current round.
• b_measured: bool, scalar, whether party B has measured in the current round.
• prev_party: Optional[str], constraint in {"a","b"} when not None, scalar, party label of the first measurer for the current round.
• prev_measurement: Optional[np.ndarray], dtype int, values in {0,1}, shape (length,), first party’s measurement vector for the current round when set.
• prev_outcome: Optional[np.ndarray], dtype int, values in {0,1}, shape (length,), first party’s outcome vector for the current round when set.

Private/internal attributes

• _rng: numpy.random.Generator, RNG used for uniform bit generation and per-index Bernoulli draws.

State model (round semantics)

• A “round” consists of up to two successful measurements, at most one by A and at most one by B, in either order; reset() clears the round state.

Planned (design-spec)

• Not specified.

Deviations

• Not specified.

Notes for Contributors

• The first successful measurement in a round stores prev_measurement and prev_outcome; the second successful measurement uses those stored arrays to generate correlated outcomes; reset() must be called to reuse the instance for a new round.
• Input validation is centralized in _validate_measurement, which coerces inputs using np.asarray(..., dtype=int); callers can pass array-like inputs as long as they coerce to a valid 1D 0/1 vector of length length.

Related

• SharedRandomness in shared_randomness.py (mentioned in the module docstring as functionally identical; not present in this module text).

Changelog

• Version 0.1: Initial implementation of PRAssisted (as indicated by the module docstring).