bayesbay.likelihood.Target

class bayesbay.likelihood.Target(name, dobs, covariance_mat_inv=None, noise_is_correlated=False, std_min=0.01, std_max=1, std_perturb_std=0.1, correlation_min=0.01, correlation_max=1, correlation_perturb_std=0.1)

Observed data with noise that can be treated as an unknown

Parameters:

  • name (str) – name of the data, for display purposes

  • dobs (np.ndarray) – numerical data

  • covariance_mat_inv (Union[Number, np.ndarray], optional) – the inverse of the data covariance matrix, either a number or a full matrix, by default None

  • noise_is_correlated (bool, optional) – whether the noise between data points is correlated or not, by default False

  • std_min (Number, optional) – the minimum value of the standard deviation of data noise, by default 0.01

  • std_max (Number, optional) – the maximum value of the standard deviation of data noise, by default 1

  • std_perturb_std (Number, optional) – the perturbation standard deviation of the standard deviation of data noise, by default 0.1

  • correlation_min (Number, optional) – the miminum value of the correlation of data noise, by default 0.01

  • correlation_max (Number, optional) – the maximum value of the correlation of data noise, by default 1

  • correlation_perturb_std (Number, optional) – the perturbation standard deviation of the standard deviation of data noise, by default 0.1

Reference Details

is_hierarchical

whether the data noise is unknown (i.e. to be inverted for)

name
initialize(state)

initializes the data noise parameters

Parameters:

state (State) – the current state in the Bayesian inference, in which DataNoiseState is to be set

inverse_covariance_times_vector(state, vector)

calculates the dot product of the covariance inverse matrix with a given vector

Parameters:

  • state (State) – the current state state

  • vector (np.ndarray) – the vector to apply the dot product on

Returns:

the result from the dot product operation

Return type:

np.ndarray

log_determinant_covariance(state)

the log of the determinant of the covariance matrix

The determinant of the data covariance matrix is calculated assuming an exponential decay in the noise correlation between adjacent data points [1], i.e.,

\[\lvert \mathbf{C}_e \rvert = \sigma^{2n} (1 - r^2)^{n-1},\]

where \(\sigma\) denotes the standard deviation, \(r\) the correlation, and \(n\) the size of the data vector.

Parameters:

state (State) – the current Bayesian inferernce state

Returns:

the log of the determinant

Return type:

float

References

back to top