<!-- filepath: /home/philip/Documents/elsciRL-Wiki/Documentation/Encoders/prior_actions_encoded.md --> # Prior Actions Encoder This encoder represents the sequence of actions taken so far in an episode as a vector or tensor. It is useful for agents that need to reason about action history (e.g., in partially observable environments). ## Class: `PriorActionsEncoder` - Maps all possible actions to indices. - Encodes the action history as a vector, preserving order. - Supports both indexed and one-hot (flattened) representations. ### Example Usage ```python from elsciRL.encoders.prior_actions_encoded import PriorActionsEncoder all_actions = ['left', 'right', 'up', 'down'] encoder = PriorActionsEncoder(all_actions) episode_action_history = ['left', 'up', 'right'] encoded = encoder.encode(episode_action_history=episode_action_history) print(encoded.shape) # Output: (output_dim,) ``` ---