v0.2.0

Release 0.2.0

Algorithms

• #41: QV-learning

Features/Enhancements

Step-size scheduling

• #35: Adds step size schedules
• #36: Adds step-size for state-action pairs

Changes

• #34: Renames learning_rate to step_size
• #36: Adds s

Release 0.0.2

This release introduces several new features:

• Double Q-learning
• Explorers: linear decay epsilon-greedy exploration and adds percentage decay epsilon-greedy exploration.
• Renames env to envs
• Adds Riverswim environment

v0.1.0

Release 0.1.0

Algorithms

• #25: SARSA

Features/Enhancements

Explorers

• #24: Expands the observe method in explorers
• #32: Adds training_mode to explorers' observe
• #31: Adds PolicyExecutor as an explorer that executes a specific policy
• #27: Provides the observation to explorers

Environments

• #26: Adds overestimation environment from Double Q-learning paper

Learners

• #28: Adds training mode to learners

Bug fixes

• #22: Fixes Transition function in River Swim Environment
• #23: Bug in explorer call in Q-learning