simrun ❭ modular_reduced_model_inference ❭ strategy ❭ _Strategy

_Strategy¶

class simrun.modular_reduced_model_inference.strategy._Strategy(name)¶

Strategy base class.

This class is used to define a strategy for the optimizer. Each strategy sets up all necessary components to define a single cost function get_score(). This cost function is used by a simrun.modular_reduced_model_inference.solver to optimize the parameters of the strategy.

Each child class must implement a _get_score class method. These are used here to construct get_score(). It is this get_score method that is optimized during optimization.

As a function of the parameters, compute a value for each trial. The optimizer will optimize for this value (highest AUROC score)

Needs some repr for input data.

E.G. A strategy that needs to optimize for AP refractory, then the Strategy needs to incorporate this data

Parameters:¶

name (str) – The name of the strategy.

Methods:¶

_get_score(x)	Compute the score for the given parameters x.
setup(data, DataSplitEvaluation)	Setup the strategy with the given data.
_setup()	Strategy-specific setup.
_get_x0()	Get an initial guess for the learnable weights of the basis functions \(\mathbf{x}\).
set_split(split, setup)	Set the split for this strategy.
get_score_static(_get_score, x, cupy_split)	static Convert the strategy-specific _get_score method to a static method.
get_y_static(y, numpy_split)	static Fetch the labels for the given split.
_objective_function_static(get_score, get_y, x)	static Compute the objective value for the given parameters x.
add_solver(solver, setup)	Add a solver to the strategy.