GaussianProcess¶

class safe_learning.GaussianProcess(gaussian_process, beta=2.0, name='gaussian_process')¶

A GaussianProcess model based on gpflow.

Parameters:	gaussian_process : instance of gpflow.models.GPModel The Gaussian process model. beta : float The scaling factor for the standard deviation to create confidence intervals.

Notes

The evaluate and gradient functions can be called with multiple arguments, in which case they are concatenated before being passed to the GP.

Attributes:	`X` Input location of observed data. `Y` Observed output. `parameters` Return the variables within the current scope. scope_name

Methods

`__call__`(self, \args, \\*kwargs)	Evaluate the function using the template to ensure variable sharing.
`add_data_point`(self, x, y)	Add data points to the GP model and update cholesky.
`build_evaluation`(self, points)	Evaluate the model, but return tensorflow tensors.
`copy_parameters`(self, other_instance)	Copy over the parameters of another instance.
`to_mean_function`(self)	Turn the uncertain function into a deterministic ‘mean’ function.
`update_feed_dict`(self)	Update the feed dictionary for tensorflow.

add_data_point(self, x, y)¶

Add data points to the GP model and update cholesky.

Parameters:	x : ndarray A 2d array with the new states to add to the GP model. Each new state is on a new row. y : ndarray A 2d array with the new measurements to add to the GP model. Each measurements is on a new row.

build_evaluation(self, points)¶: Evaluate the model, but return tensorflow tensors.

copy_parameters(self, other_instance)¶: Copy over the parameters of another instance.

to_mean_function(self)¶: Turn the uncertain function into a deterministic ‘mean’ function.