import numpy as np
import pandas as pd
from qutip import qfunc
import warnings
from qsttoolkit.data.state_batches import FockStates, CoherentStates, ThermalStates, NumStates, BinomialStates, CatStates, GKPStates, RandomStates
from qsttoolkit.data.noise import mixed_state_noise, gaussian_convolution, apply_measurement_noise
[docs]
def optical_state_dataset(dim: int, data_dim: int=None, state_numbers: list=[1000, 1000, 1000, 1000, 1000, 1000, 1000, 1000], mixed_state_noise_level: float=0.2, Gaussian_conv_ntherm: float=2.0, affine_theta: float=20.0, affine_x: float=0.1, affine_y: float=0.1, additive_Gaussian_stddev: float=0.001, pepper_p: float=0.01, salt_p: float=0.0, latent_dim=None, mixed_state_noise_noise_level=None) -> pd.DataFrame:
"""
Generates a standardized dataset of optical quantum states with added noise for training machine learning quantum state discrimination and tomography models.
Parameters
----------
dim : int
Dimensionality of the Hilbert space.
data_dim : int
Number of points in the X and P grids for the Q-function.
state_numbers : list[int]
List of 8 integers specifying the number of states to generate for each state type in the dataset. order is [Fock, Coherent, Thermal, Num, Binomial, Cat, GKP, Random]. Defalts to [1000, 1000, 1000, 1000, 1000, 1000, 1000, 1000].
mixed_state_noise_level : float
Random state coefficient for the mixed state noise. Defaults to 0.2.
Gaussian_conv_ntherm : float
Mean photon number, equal to the variance of the Gaussian convolution kernel. Defaults to 2.0.
affine_theta : float
Maximum rotation angle in degrees. Defaults to 20.0.
affine_x : float
Maximum translation in the x direction. Defaults to 0.1.
affine_y : float
Maximum translation in the y direction. Defaults to 0.1.
additive_Gaussian_stddev : float
Standard deviation of the Gaussian distribution from which additive noise is sampled. Defaults to 0.001.
salt_p : float
Proportion of pixels to set to 1. Defaults to 0.0.
pepper_p : float
Proportion of pixels to set to 0. Defaults to 0.01.
Returns
-------
pd.DataFrame
DataFrame containing the labels, density matrices, Husimi Q functions, and state parameters of the states.
"""
if not data_dim:
if latent_dim:
data_dim = latent_dim
warnings.warn("latent_dim is deprecated and will be removed in a future version. Please use data_dim instead.", DeprecationWarning, stacklevel=2)
else:
raise ValueError("data_dim must be specified.")
if not mixed_state_noise_level:
if mixed_state_noise_noise_level:
mixed_state_noise_level = mixed_state_noise_noise_level
warnings.warn("mixed_state_noise_noise_level is deprecated and will be removed in a future version. Please use mixed_state_noise_level instead.", DeprecationWarning, stacklevel=2)
else:
raise ValueError("mixed_state_noise_level must be specified.")
if not isinstance(dim, int):
raise ValueError("dim must be an integer.")
if not isinstance(data_dim, int):
raise ValueError("data_dim must be an integer.")
if not (isinstance(state_numbers, list) and len(state_numbers) == 8 and all(isinstance(n, int) and n >= 0 for n in state_numbers)):
raise ValueError("state_numbers must be a list of 8 positive integers.")
fock_batch = FockStates(n_states = state_numbers[0],
dim = dim,
n_range = [0, dim])
coherent_batch = CoherentStates(n_states = state_numbers[1],
dim = dim,
alpha_magnitude_range = [1e-6, 3])
thermal_batch = ThermalStates(n_states = state_numbers[2],
dim = dim,
nbar_range = [0, dim])
num_batch = NumStates(n_states = state_numbers[3],
dim = dim,
types = ['17', 'M', 'P', 'P2', 'M2'])
binomial_batch = BinomialStates(n_states = state_numbers[4],
dim = dim,
S_range = [1, 10],
mu_range = [0, 2])
cat_batch = CatStates(n_states = state_numbers[5],
dim = dim,
alpha_magnitude_range = [0, 10])
gkp_batch = GKPStates(n_states = state_numbers[6],
dim = dim,
n1_range = [-20, 20],
n2_range = [-20, 20],
delta_range=[0.2, 0.5],
mu_range=[0, 2])
random_batch = RandomStates(n_states = state_numbers[7],
dim = dim)
print('States generated')
# Create phase space grid
xgrid = np.linspace(-5, 5, data_dim)
pgrid = np.linspace(-5, 5, data_dim)
fock_data = pd.DataFrame(columns=['label', 'density_matrix', 'Husimi-Q_function', 'state_parameter'])
coherent_data = pd.DataFrame(columns=['label', 'density_matrix', 'Husimi-Q_function', 'state_parameter'])
thermal_data = pd.DataFrame(columns=['label', 'density_matrix', 'Husimi-Q_function', 'state_parameter'])
num_data = pd.DataFrame(columns=['label', 'density_matrix', 'Husimi-Q_function', 'state_parameter'])
binomial_data = pd.DataFrame(columns=['label', 'density_matrix', 'Husimi-Q_function', 'state_parameter'])
cat_data = pd.DataFrame(columns=['label', 'density_matrix', 'Husimi-Q_function', 'state_parameter'])
gkp_data = pd.DataFrame(columns=['label', 'density_matrix', 'Husimi-Q_function', 'state_parameter'])
random_data = pd.DataFrame(columns=['label', 'density_matrix', 'Husimi-Q_function', 'state_parameter'])
warnings.warn("The column 'Husimi-Q' will be renamed to 'Husimi_Q' in the next major update.", FutureWarning, stacklevel=2)
print("DataFrames initialized")
fock_densities = [mixed_state_noise(dm, mixed_state_noise_level) for dm in fock_batch.density_matrices()]
fock_data['label'] = ['fock']*len(fock_densities)
fock_data['density_matrix'] = [dm.full() for dm in fock_densities]
fock_data['Husimi-Q_function'] = [apply_measurement_noise(gaussian_convolution(qfunc(dm, xgrid, pgrid), Gaussian_conv_ntherm), affine_theta, affine_x, affine_y, additive_Gaussian_stddev, pepper_p, salt_p) for dm in fock_densities]
fock_data['state_parameter'] = fock_batch.params
print("Fock data generated")
coherent_densities = [mixed_state_noise(dm, mixed_state_noise_level) for dm in coherent_batch.density_matrices()]
coherent_data['label'] = ['coherent']*len(coherent_densities)
coherent_data['density_matrix'] = [dm.full() for dm in coherent_densities]
coherent_data['Husimi-Q_function'] = [apply_measurement_noise(gaussian_convolution(qfunc(dm, xgrid, pgrid), Gaussian_conv_ntherm), affine_theta, affine_x, affine_y, additive_Gaussian_stddev, pepper_p, salt_p) for dm in coherent_densities]
coherent_data['state_parameter'] = coherent_batch.params
print("Coherent data generated")
thermal_densities = [mixed_state_noise(dm, mixed_state_noise_level) for dm in thermal_batch.density_matrices()]
thermal_data['label'] = ['thermal']*len(thermal_densities)
thermal_data['density_matrix'] = [dm.full() for dm in thermal_densities]
thermal_data['Husimi-Q_function'] = [apply_measurement_noise(gaussian_convolution(qfunc(dm, xgrid, pgrid), Gaussian_conv_ntherm), affine_theta, affine_x, affine_y, additive_Gaussian_stddev, pepper_p, salt_p) for dm in thermal_densities]
thermal_data['state_parameter'] = thermal_batch.params
print("Thermal data generated")
num_densities = [mixed_state_noise(dm, mixed_state_noise_level) for dm in num_batch.density_matrices()]
num_data['label'] = ['num']*len(num_densities)
num_data['density_matrix'] = [dm.full() for dm in num_densities]
num_data['Husimi-Q_function'] = [apply_measurement_noise(gaussian_convolution(qfunc(dm, xgrid, pgrid), Gaussian_conv_ntherm), affine_theta, affine_x, affine_y, additive_Gaussian_stddev, pepper_p, salt_p) for dm in num_densities]
num_data['state_parameter'] = num_batch.params
print("Num data generated")
binomial_densities = [mixed_state_noise(dm, mixed_state_noise_level) for dm in binomial_batch.density_matrices()]
binomial_data['label'] = ['binomial']*len(binomial_densities)
binomial_data['density_matrix'] = [dm.full() for dm in binomial_densities]
binomial_data['Husimi-Q_function'] = [apply_measurement_noise(gaussian_convolution(qfunc(dm, xgrid, pgrid), Gaussian_conv_ntherm), affine_theta, affine_x, affine_y, additive_Gaussian_stddev, pepper_p, salt_p) for dm in binomial_densities]
binomial_data['state_parameter'] = binomial_batch.params
print("Binomial data generated")
cat_densities = [mixed_state_noise(dm, mixed_state_noise_level) for dm in cat_batch.density_matrices()]
cat_data['label'] = ['cat']*len(cat_densities)
cat_data['density_matrix'] = [dm.full() for dm in cat_densities]
cat_data['Husimi-Q_function'] = [apply_measurement_noise(gaussian_convolution(qfunc(dm, xgrid, pgrid), Gaussian_conv_ntherm), affine_theta, affine_x, affine_y, additive_Gaussian_stddev, pepper_p, salt_p) for dm in cat_densities]
cat_data['state_parameter'] = cat_batch.params
print("Cat data generated")
gkp_densities = [mixed_state_noise(dm, mixed_state_noise_level) for dm in gkp_batch.density_matrices()]
gkp_data['label'] = ['gkp']*len(gkp_densities)
gkp_data['density_matrix'] = [dm.full() for dm in gkp_densities]
gkp_data['Husimi-Q_function'] = [apply_measurement_noise(gaussian_convolution(qfunc(dm, xgrid, pgrid), Gaussian_conv_ntherm), affine_theta, affine_x, affine_y, additive_Gaussian_stddev, pepper_p, salt_p) for dm in gkp_densities]
gkp_data['state_parameter'] = gkp_batch.params
print("GKP data generated")
random_densities = [mixed_state_noise(dm, mixed_state_noise_level) for dm in random_batch.density_matrices()]
random_data['label'] = ['random']*len(random_densities)
random_data['density_matrix'] = [dm.full() for dm in random_densities]
random_data['Husimi-Q_function'] = [apply_measurement_noise(gaussian_convolution(qfunc(dm, xgrid, pgrid), Gaussian_conv_ntherm), affine_theta, affine_x, affine_y, additive_Gaussian_stddev, pepper_p, salt_p) for dm in random_densities]
random_data['state_parameter'] = random_batch.params
print("Random data generated")
data = pd.concat([fock_data, coherent_data, thermal_data, num_data, binomial_data, cat_data, gkp_data, random_data])
data = data.sample(frac=1).reset_index(drop=True)
print("Dataset generated")
return data