figaro.transform module

figaro.transform.gradient_inv_jacobian(x, bounds, flag=True)

logarithmic gradient of the probit transformation Jacobian

Parameters:

• x (np.ndarray) – sample(s) to evaluate the jacobian at

• bounds (np.ndarray) – limits for each dimension (2d array, [[xmin, xmax], [ymin, ymax]…])

• flag (bool) – whether to skip the evaluation

Returns:

log jacobian (ones if flag is False)

Return type:

np.ndarray

figaro.transform.log2PI = 1.8378770664093453

//docs.scipy.org/doc/scipy/reference/generated/scipy.special.erf.html See notes there for cumulative of the unit normal distribution.

Type:

Scipy’s implementation of ERF

Type:

https

figaro.transform.probit_logJ(x, bounds, flag=True)

Jacobian of the probit transformation marginalised over dimensions

Parameters:

• x (np.ndarray) – sample(s) to evaluate the jacobian at

• bounds (np.ndarray) – limits for each dimension (2d array, [[xmin, xmax], [ymin, ymax]…])

• flag (bool) – whether to skip the evaluation

Returns:

log jacobian (zeros if flag is False)

Return type:

np.ndarray

figaro.transform.probit_log_jacobian(x, bounds, flag=True)

Jacobian of the probit transformation

Parameters:

• x (np.ndarray) – sample(s) to evaluate the jacobian at

• bounds (np.ndarray) – limits for each dimension (2d array, [[xmin, xmax], [ymin, ymax]…])

• flag (bool) – whether to skip the evaluation

Returns:

log jacobian (zeros if flag is False)

Return type:

np.ndarray

figaro.transform.transform_from_probit(x, bounds)

Coordinate change from probit to natural space. cdf_normal is the cumulative distribution function of the unit normal distribution.

x(t) = xmin + (xmax-xmin)*cdf_normal(t|0,1)

Parameters:

• x (np.ndarray) – sample(s) to antitransform (2d array)

• bounds (np.ndarray) – limits for each dimension (2d array, [[xmin, xmax], [ymin, ymax]…])

Returns:

sample(s)

Return type:

np.ndarray

figaro.transform.transform_to_probit(x, bounds)

Coordinate change into probit space. cdf_normal is the cumulative distribution function of the unit normal distribution. WARNING: returns NAN if x is not in [xmin, xmax].

t(x) = cdf^-1_normal((x-x_min)/(x_max - x_min))

Parameters:

• x (np.ndarray) – sample(s) to transform (2d array)

• bounds (np.ndarray) – limits for each dimension (2d array, [[xmin, xmax], [ymin, ymax]…])

Returns:

sample(s)

Return type:

np.ndarray