rlportfolio.policy.gpm module

class GPM

Bases: Module

__init__(edge_index: ndarray | Tensor, edge_type: ndarray, nodes_to_select: ndarray | list[int] | Tensor, initial_features: int = 3, k_short: int = 3, k_medium: int = 21, conv_mid_features: int = 3, conv_final_features: int = 20, graph_layers: int = 1, time_window: int = 50, softmax_temperature: float = 1, num_workers: int = 1, device: str = 'cpu') → GPM

GPM (Graph-based Portfolio Management) policy network initializer.

Parameters:

• edge_index – Graph connectivity in COO format.

• edge_type – Type of each edge in edge_index.

• nodes_to_select – ID of nodes to be selected to the portfolio.

• initial_features – Number of input features.

• k_short – Size of short convolutional kernel.

• k_medium – Size of medium convolutional kernel.

• conv_mid_features – Size of intermediate convolutional channels.

• conv_final_features – Size of final convolutional channels.

• graph_layers – Number of graph neural network layers.

• time_window – Size of time window used as agent’s state.

• softmax_temperature – Temperature parameter to softmax function.

• num_workers – Number of workers to use in parallel execution of graph batches.

• device – Device in which the neural network will be run.

Note

Reference article: https://doi.org/10.1016/j.neucom.2022.04.105.

forward(observation: Tensor, last_action: Tensor) → Tensor

Policy network’s forward propagation. Defines a most favorable action of this policy given the inputs.

Parameters:

• observation – environment observation.

• last_action – Last action performed by agent.

Returns:

Action to be taken.

class GPMSimplified

Bases: GPM

__init__(edge_index: ndarray | Tensor, edge_type: ndarray, nodes_to_select: ndarray | list[int] | Tensor, initial_features: int = 3, k_short: int = 3, k_medium: int = 21, conv_mid_features: int = 3, conv_final_features: int = 20, graph_layers: int = 1, time_window: int = 50, softmax_temperature: float = 1, num_workers: int = 1, device: str = 'cpu') → GPMSimplified

GPM (Graph-based Portfolio Management) policy network initializer.

Parameters:

• edge_index – Graph connectivity in COO format.

• edge_type – Type of each edge in edge_index.

• nodes_to_select – ID of nodes to be selected to the portfolio.

• initial_features – Number of input features.

• k_short – Size of short convolutional kernel.

• k_medium – Size of medium convolutional kernel.

• conv_mid_features – Size of intermediate convolutional channels.

• conv_final_features – Size of final convolutional channels.

• graph_layers – Number of graph neural network layers.

• time_window – Size of time window used as agent’s state.

• softmax_temperature – Temperature parameter to softmax function.

• num_workers – Number of workers to use in parallel execution of graph batches.

• device – Device in which the neural network will be run.

Note

Reference article: https://doi.org/10.1016/j.neucom.2022.04.105.

forward(observation: Tensor, last_action: Tensor) → Tensor

Policy network’s forward propagation. Defines a most favorable action of this policy given the inputs.

Parameters:

• observation – environment observation.

• last_action – Last action performed by agent.

Returns:

Action to be taken.