from .optimiser import LocalOptimiser
__all__ = ['GradientDescent']
[docs]
class GradientDescent(LocalOptimiser):
"""
Implementation of a gradient descent update.
Parameters
----------
variable : Variable
Variable to which the optimiser refers.
step : float, optional
Step size for the update, defaults to 1.
kwargs
Extra parameters to be used by the class.
"""
def __init__(self, variable, **kwargs):
super().__init__(variable, **kwargs)
self.step_size = kwargs.pop('step_size', 1.)
[docs]
async def step(self, step_size=None, grad=None, processed_grad=None, **kwargs):
"""
Apply the optimiser.
Parameters
----------
step_size : float, optional
Step size to use for this application, defaults to instance step.
grad : Data, optional
Gradient to use for the step, defaults to variable gradient.
processed_grad : Data, optional
Processed gradient to use for the step, defaults to processed variable gradient.
kwargs
Extra parameters to be used by the method.
Returns
-------
Variable
Updated variable.
"""
step_size = self.step_size if step_size is None else step_size
processed_grad = await self.pre_process(grad=grad,
processed_grad=processed_grad,
**kwargs)
direction = processed_grad
self.variable.data[:] -= step_size*direction.data
await self.post_process(**kwargs)
return self.variable
[docs]
def reset(self, **kwargs):
"""
Reset optimiser state along with any stored buffers.
Parameters
----------
kwargs
Extra parameters to be used by the method.
Returns
-------
"""
pass