autonomous_calibration_engine.AdaptivePriorEvolution Class Reference

Public Member Functions
	__init__ (self, int n_params, float forgetting_factor=0.995)
	update_from_sensor (self, int sensor_id, np.ndarray posterior_mean, np.ndarray posterior_cov, float quality)
Tuple[np.ndarray, np.ndarray, float]	get_prior (self)
Tuple[np.ndarray, np.ndarray]	predict_future_drift (self, float time_horizon)

Public Attributes
	n_params = n_params
	forgetting_factor = forgetting_factor
tuple	prior_mean = np.zeros(n_params)
tuple	prior_covariance
dict	sensor_posteriors = {}
int	update_count = 0
float	effective_sample_size = 1.0
deque	history = deque(maxlen=100)
	drift_detector = DriftDetector(n_params)

Protected Member Functions
	_empirical_bayes_update (self)
float	_compute_confidence (self)

Detailed Description

Evolves population prior based on ongoing system performance.
Uses empirical Bayes to continuously improve prior from all sensors.

Definition at line 56 of file autonomous_calibration_engine.py.

Constructor & Destructor Documentation

__init__()

autonomous_calibration_engine.AdaptivePriorEvolution.__init__	(		self,
		int	n_params,
		float	forgetting_factor = 0.995 )

Definition at line 62 of file autonomous_calibration_engine.py.

Member Function Documentation

_compute_confidence()

float autonomous_calibration_engine.AdaptivePriorEvolution._compute_confidence ( self )

protected

Compute confidence in current prior

Definition at line 183 of file autonomous_calibration_engine.py.

Referenced by update_from_sensor().

_empirical_bayes_update()

autonomous_calibration_engine.AdaptivePriorEvolution._empirical_bayes_update ( self )

protected

Empirical Bayes: Estimate population prior from sensor posteriors.

Model: θⱼ ~ N(μ_pop, Σ_pop)  (sensors drawn from population)

Estimate μ_pop = mean(θⱼ) and Σ_pop = var(θⱼ) + mean(Σⱼ)

Definition at line 132 of file autonomous_calibration_engine.py.

References effective_sample_size, n_params, prior_covariance, prior_mean, and sensor_posteriors.

Referenced by update_from_sensor().

get_prior()

Tuple[np.ndarray, np.ndarray, float] autonomous_calibration_engine.AdaptivePriorEvolution.get_prior

(

self

)

Get current prior (mean, covariance, confidence)

Definition at line 200 of file autonomous_calibration_engine.py.

predict_future_drift()

Tuple[np.ndarray, np.ndarray] autonomous_calibration_engine.AdaptivePriorEvolution.predict_future_drift	(		self,
		float	time_horizon )

Predict how parameters will drift over time horizon.

Returns: (predicted_mean, predicted_covariance)

Definition at line 208 of file autonomous_calibration_engine.py.

update_from_sensor()

autonomous_calibration_engine.AdaptivePriorEvolution.update_from_sensor	(		self,
		int	sensor_id,
		np.ndarray	posterior_mean,
		np.ndarray	posterior_cov,
		float	quality )

Update population prior from individual sensor posterior.
Uses empirical Bayes: prior is the mean of all sensor posteriors.

Definition at line 86 of file autonomous_calibration_engine.py.

References _compute_confidence(), _empirical_bayes_update(), drift_detector, effective_sample_size, forgetting_factor, history, feedtwin.session.core.Session.history, prior_covariance, prior_mean, sensor_posteriors, time.time(), and update_count.

Member Data Documentation

drift_detector

autonomous_calibration_engine.AdaptivePriorEvolution.drift_detector = DriftDetector(n_params)

Definition at line 80 of file autonomous_calibration_engine.py.

Referenced by update_from_sensor().

effective_sample_size

float autonomous_calibration_engine.AdaptivePriorEvolution.effective_sample_size = 1.0

Definition at line 76 of file autonomous_calibration_engine.py.

Referenced by _empirical_bayes_update(), and update_from_sensor().

forgetting_factor

autonomous_calibration_engine.AdaptivePriorEvolution.forgetting_factor = forgetting_factor

Definition at line 64 of file autonomous_calibration_engine.py.

Referenced by robust_calibration.RobustCalibrationFramework.add_calibration_point(), robust_calibration.RobustCalibrationFramework.get_calibration_summary(), robust_calibration.RobustCalibrationFramework.recursive_least_squares_update(), and update_from_sensor().

history

deque autonomous_calibration_engine.AdaptivePriorEvolution.history = deque(maxlen=100)

Definition at line 79 of file autonomous_calibration_engine.py.

Referenced by update_from_sensor().

n_params

autonomous_calibration_engine.AdaptivePriorEvolution.n_params = n_params

Definition at line 63 of file autonomous_calibration_engine.py.

Referenced by _empirical_bayes_update().

prior_covariance

tuple autonomous_calibration_engine.AdaptivePriorEvolution.prior_covariance

Initial value:

=  np.diag(
            [10000.0, 100000.0] + [10000.0] * (n_params - 2)
        )

Definition at line 69 of file autonomous_calibration_engine.py.

Referenced by _empirical_bayes_update(), and update_from_sensor().

prior_mean

tuple autonomous_calibration_engine.AdaptivePriorEvolution.prior_mean = np.zeros(n_params)

Definition at line 67 of file autonomous_calibration_engine.py.

Referenced by _empirical_bayes_update(), and update_from_sensor().

sensor_posteriors

autonomous_calibration_engine.AdaptivePriorEvolution.sensor_posteriors = {}

Definition at line 74 of file autonomous_calibration_engine.py.

Referenced by _empirical_bayes_update(), and update_from_sensor().

update_count

int autonomous_calibration_engine.AdaptivePriorEvolution.update_count = 0

Definition at line 75 of file autonomous_calibration_engine.py.

Referenced by update_from_sensor().

---

The documentation for this class was generated from the following file:

• daq-server/tools/calibration/autonomous_calibration_engine.py