Adaptive Engine

class tsuchinoko.adaptive.Engine

The Adaptive Engine base class. This component is generally to be responsible for determining future measurement targets.

abstract request_targets(position: Tuple) → Iterable[Tuple]

Determine new targets to be measured

Parameters:

position: tuple

The current ‘position’ of the experiment in the target domain.

Returns:

targets: array_like

The new targets to be measured.

abstract reset()

Called when an experiment stops, or is about to start. Returns the engine to a clean state.

abstract train()

Perform training. This can be short-circuited to only train on every N-th iteration, for example.

abstract update_measurements(data: Data)

Update internal variables with the provided new data

abstract update_metrics(data: Data)

Calculates various metrics to drive visualizations for the client. The data object is expected to be mutated to include these new values.