ribs.emitters.opt.GradientOptBase¶

class ribs.emitters.opt.GradientOptBase(theta0: ArrayLike, lr: float | floating)[source]¶

Base class for gradient-based optimizers.

Note

These optimizers are designed for gradient ascent rather than gradient descent.

These optimizers maintain a current solution point \(\theta\). The solution point is obtained with the theta property, and it is updated by passing a gradient to step(). Finally, the point can be reset to a new value with reset().

Your constructor may take in additional arguments beyond theta0 and lr, but expect that these two arguments will always be passed in.

Parameters:¶

theta0: ArrayLike¶

Initial solution. 1D array.

lr: float | floating¶

Learning rate for the update.

Methods

reset(theta0)	Resets the solution point to a new value.
step(gradient)	Ascends the solution based on the given gradient.

Attributes

theta

The current solution point.

abstract reset(theta0: ArrayLike) → None[source]¶

Resets the solution point to a new value.

Parameters:¶

theta0: ArrayLike¶

The new solution point. 1D array.

abstract step(gradient: ArrayLike) → None[source]¶

Ascends the solution based on the given gradient.

Parameters:¶

gradient: ArrayLike¶

The (estimated) gradient of the current solution point. 1D array.

abstract property theta : ndarray¶

The current solution point.