pyriemann_qiskit.classification.QuantumStateDiscriminator¶
- class pyriemann_qiskit.classification.QuantumStateDiscriminator(n_jobs=1)[source]¶
Quantum state classifier using the Pretty Good Measurement (PGM).
The mental state of the user (class A or B) is modeled as a mixed quantum state (density matrix) rho_c, estimated from training EEG via quantum state tomography. Class priors pi_c are estimated from class frequencies in the training set.
The classifier is a POVM (Positive Operator-Valued Measure) built via the Pretty Good Measurement:
Pi_c = rho_total^{-1/2} (pi_c * rho_c) rho_total^{-1/2}
where rho_total = sum_c pi_c * rho_c is the prior-weighted average state.
The POVM satisfies sum_c Pi_c = I, so scores trace(Pi_c . M) are valid probabilities (non-negative, summing to 1) directly from the Born rule — no softmax needed.
For two classes with equal priors, this approximates the Helstrom measurement (theoretically optimal quantum state discrimination).
- Parameters:
n_jobs (int, default=1) – Number of parallel jobs over trials at predict time.
- density_matrices_¶
Per-class density matrices rho_c (trace=1, PSD), estimated via quantum state tomography.
- Type:
dict[label -> ndarray (n_channels, n_channels)]
- povm_¶
Per-class POVM elements Pi_c satisfying sum_c Pi_c = I.
- Type:
dict[label -> ndarray (n_channels, n_channels)]
- priors_¶
Class prior probabilities estimated from training set frequencies.
- Type:
dict[label -> float]
- classes_¶
Unique class labels seen at fit time.
- Type:
ndarray
Notes
Added in version 0.5.0.
Changed in version 0.6.0: Moved to algorithms sub-package