The function of epsilon in this algorithm

[su123123123]

I rebuild the algorithm,but at the "self.boosted_layers[i] = torch.from_numpy(np.array(self.xg.fit(x0.detach().numpy(), (self.l).detach().numpy()).feature_importances_) + self.epsilon)", I can not understand , that feature_importances_ will not change . I think that the function of epsilon does not work.

[tusharsarkar3]

Epsilon comes to the picture when the importance of a feature is 0. We take the log of feature importance in the later steps of the algorithm to maintain the same order of the weights and we want to avoid 1 as the order when we have 0 importance of the feature so we have used epsilon. Does this solve your issue ?

[su123123123]

Thank you,sir.I got it.

…

------------------ 原始邮件 ------------------ 发件人: "Tushar ***@***.***>; 发送时间: 2021年6月20日(星期天) 晚上8:28 收件人: ***@***.***>; 抄送: ***@***.***>; ***@***.***>; 主题: Re: [tusharsarkar3/XBNet] The function of epsilon in this algorithm (#2) Epsilon comes to the picture when the importance of a feature is 0. We take the log of feature importance in the later steps of the algorithm to maintain the same order of the weights and we want to avoid 1 as the order when we have 0 importance of the feature so we have used epsilon. Does this solve your issue ? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.