mixing matrix of unseen data?

[ahassaine]

Hi there,

Just wondering if there is a way in nimfa to obtain the mixing matrix of unseen data?
Equivalent of transform in sklearn.decomposition ?

Thanks!

[davek44]

I'm also interested in this functionality.

[jkapila]

I wish that was available. This package could have been greatly utilised.
Folks please help us on this.

[alailink]

I've been looking into this for a few days. Here's a link to how to do it manually, from a similar issue:

#43

I agree this should be easy to implement package-wide as a standard feature, similar to sklearn transform.
However, without minor adjustments it looks like you could do something like

nmf= nimfa.Nmf(V)
fit = nmf()
weights = fit.coef()
nimfa.Nmf(V2, H=weights, update=None)

the catch being, update does not allow None as currently written, but it should be easy to change in your local code.

if I have time I may pull this and attempt to implement. no promises :(

[Xuanzewuyan]

我已经研究了几天了。这是来自类似问题的如何手动执行此操作的链接：

#43

我同意这应该很容易作为标准功能在包范围内实现，类似于 sklearn 转换。但是，如果没有细微的调整，看起来您可以做类似的事情

nmf= nimfa.Nmf(V)
fit = nmf()
weights = fit.coef()
nimfa.Nmf(V2, H=weights, update=None)

问题是，更新不允许当前编写的 None ，但应该很容易更改本地代码。

如果我有时间，我可以拉这个并尝试实施。没有承诺:(

i have already done this.I modify the source code, through deliver the parameter, now you can only update W（or H），other one will fiexed. If anyone needs help, please let me know. My email is [email protected]