clamp cross-entropy loss

3 次查看（过去 30 天）
Matt Fetterman
Matt Fetterman 2020-9-3
评论： Matt Fetterman 2020-9-6
the Matlab cross-entropy loss has this form:
loss = -sum(W*(T.*log(Y)))/N;
I would like to "clamp" it so that the log function output is bounded, for example it cannot be less than 100.
See here: https://pytorch.org/docs/stable/generated/torch.nn.BCELoss.html
Can we do it?

请先登录，再回答此问题。

采纳的回答

David Goodmanson
David Goodmanson 2020-9-3
编辑：David Goodmanson 2020-9-6
Hi Matt,
z = log(Y);
z(z<100) = 100;
loss = -sum(W*(T.*z))/N;
In the link you provided, they talk about a limit of -100 rather than +100. The former appears to make more sense. Lots of possibilities for a smooth differentiable cutoff, here is one, assuming Y>=0
Ylimit = -100;
loss = -sum(W*(T.*log(Y+exp(Ylimit)))/N;
  3 个评论
显示 1更早的评论
David Goodmanson
David Goodmanson 2020-9-6
Hi Matt,
see amended answer.
Matt Fetterman
Matt Fetterman 2020-9-6
probably a smart approach.

请先登录，再进行评论。

更多回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Matrix Indexing 的更多信息

标签

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by