optimizer object
optimizer object
Perform a gradient-based optimization step
Perform a gradient-based optimization step
Data to fit the model
Minibatch size for each iteration in the ascent procedure.
Minibatch size for each iteration in the ascent procedure. If None, it performs full-batch optimization
convergence flag for local optimisation
convergence flag for local optimisation
Error tolerance in log-likelihood for the stopping criteria
Error tolerance in log-likelihood for the stopping criteria
this prevents the seed from repeating every time step() is called which would cause the same samples being taken
this prevents the seed from repeating every time step() is called which would cause the same samples being taken
Current loss value
Current loss value
Maximum number of iterations allowed
Maximum number of iterations allowed
Optional regularization term
Optional regularization term
random seed for mini-batch sampling
random seed for mini-batch sampling
Update model parameters using streaming data
Update model parameters using streaming data
Streaming data
Perform a gradient-based optimization step
Perform a gradient-based optimization step
Data to fit the model
Linear Algebra operations necessary for computing updates for the parameters
Linear Algebra operations necessary for computing updates for the parameters
This is to avoid duplicating code for Gaussian and Weights updates in the optimization algorithms' classes
Contains basic functionality for an object that can be modified by Optimizer