malt.agents.player.ModelBasedPlayer

class malt.agents.player.ModelBasedPlayer(model: malt.models.supervised_model.SupervisedModel, policy: Callable, trainer: Callable, merchant: malt.agents.merchant.Merchant, assayer: malt.agents.assayer.Assayer, portfolio: Optional[malt.data.dataset.Dataset] = None)[source]

Bases: malt.agents.player.Player

Player with model.

Parameters
  • model (SupervisedModel) – Model to predict properties based on structure.

  • policy (Callable) – Policy to rank candidates.

  • trainer (Callable) – Function to train a model.

  • merchant (Merchant) – Merchant that merchandizes candidates.

  • assayer (Assayer) – Assayer that assays the candidates.

  • portfolio (Dataset) – Initial knowledge about data points.

Note

  1. Portfolio respects order and could be used to analyze acquisition

    trajectory.

__init__(model: malt.models.supervised_model.SupervisedModel, policy: Callable, trainer: Callable, merchant: malt.agents.merchant.Merchant, assayer: malt.agents.assayer.Assayer, portfolio: Optional[malt.data.dataset.Dataset] = None)[source]

Methods

__init__(model, policy, trainer, merchant, ...)

assay(dataset)

merchandize(dataset)

prioritize()

step(*args, **kwargs)

train()