About sharpe in train_AFF.py

[wangtianqi1993]

Hello, in line 115 of train_AFF.py, the code is:
sharpe = ((ret_mean - 0.03/252) / ret.std() * np.sqrt(252)).item()
It calculates the Sharpe ratio using ret. However, ret should be the correlation between fct and tgt. I don't understand why the Sharpe ratio can be calculated using ret in this context.

[silentsky2billten]

pls see the following formula :

[shengcai-win]

I also found this problem, but I don't know what ret is. From the code, I feel that it is neither a correlation coefficient nor a rate of return. I don't know why it can be used to calculate sharpe.

[DulyHao]

Thank you for your issue, sharpe ratio here is not used in factor mining.
That should be a deprecated code as we tried to process time-series task instead of cross-sectional (this paper). You may ignore that.

[shengcai-win]

Thanks, could you please tell me what ret is? In your code, ret is used to calculate the correlation of each factor, but normally, the correlation of factors should be calculated using factor values.