Added
- Added class `TradingEnvXY`, a high-level implementation of `TradingEnv` that
allows to pass features in tabular form (X) and prices for assets to be
traded (Y) as pandas.DataFrame.
- Added class `EventNewObservation` to `Transmitter` to notify the environment
that a new observation is available. This makes it easier to create custom
environments whose observations can be expressed in tabular form.
- Reward class `LogReturn` now supports extra arguments to facilitate the
scaling and clipping of the reward within a predefined range.
- Added class `State`, an implementation of `IState` which makes it easier to
get started with use cases where observations can be expressed in tabular
form.
- Added attribute `TradingEnv.visits` to keep track of the number of visits to
each state in the environment during training.
- Added optional argument `episode_length` to `TradingEnv` to specify the
maximum number of steps in an episode. This is useful for environments where
the default episode length is too large.
- Added docstrings and documentation.
Changed
- Refactored contracts in `tradingenv.contract` module.