AI4Finance-Foundation / AI4Finance-Foundation/FinRL
Reward function in StockTradingEnvStopLoss
- 主要語言
- Jupyter Notebook
- 星號
- 16.3k
- 分支
- 3.5k
- PR 合併指標
- 30 天內沒有已合併 PR
描述
Referencing the env_stocktrading_stoploss.py , it states that "Reward at any step is given as follows
**r_i = (sum(cash, asset_value) + additional_reward - total_penalty - initial_cash) / initial_cash / days_elapsed"**.
But if we look at how reward is defined under get_reward(self) [lines 282-284], days elapsed seemed to be omitted.
**reward = (
(total_assets - total_penalty + additional_reward) / self.initial_amount
) - 1**
Is there a reason for this or there is an implicit assumption taken?
https://github.com/AI4Finance-Foundation/FinRL/blob/master/finrl/finrl_meta/env_stock_trading/env_stocktrading_stoploss.py
貢獻指南
這個儲存庫沒有索引到貢獻指南
評估
這個 Issue 還沒有評估資料。