However, recent changes in Google Groups disabled the view for people
not being members of those groups yet. That's fixed now. Since I
usually post to the groups via email and not through the embedded web
interface on the ADAMS homepage, I didn't notice this regression.
> Installed latest, stable version of ADAMS with all add-ons. From the
> 'Machine Learning' menu, I am able to launch all applications (MEKA, Weka
> Explorer, Experimenter, etc.), but not MOA.
> Is there some config setting that I must define to get this menu-item to
No, it was a library incompatibility between the jclasslocator library
that MOA used (0.0.12) and the version that ADAMS uses (0.0.15). I've
deployed a custom (unofficial) version of MOA that fixes that problem
and kicked off an ADAMS build. Download a snapshot from the ADAMS
homepage in about an hour. Subsequent releases of MOA will have that
fix as well.
> I can confirm it is now working, but I have stumbled on another Weka/MOA
> issue, I hope you can assist with:
> In the Weka Explorer (launched from the ADAMS menu and also the standalone
> 3.9.4 version), I cannot view/select MOA resources (generators, classifiers,
> etc.). Works with standalone Weka 3.8.4.
> Included below is my environment for reference.
Please try again. If you get exceptions on the command-line, please
post those as they help tracking down the where the problem occurs.
BTW I don't use MOA, so I don't notice such regressions unfortunately.
For other questions regarding ADAMS, please use the ADAMS forums.
I'm trying to fuse together two machine learning systems, one supervised, the other unsupervised. Both are multi-class systems and so far they've both been predicting the same range of classes (a nominal class attribute).
However, the actual problem requires more classes to be added for which I don't have training data. This means the supervised part of the system will never predict these, while the unsupervised part can handle it perfectly well (albeit with lower accuracy).
As a first step I have changed the header of my nominal class attribute to include all possible classes, even those for which I don't have training data. My question is: how are supervised classifiers from WEKA actually handling such untrained classes? What would the resulting likelihood distribution look like? Would it contain zero's for unchanged classes? Or chance-level values for them? I think the latter would be ideal for my purposes...
ps. I tried posting this message yesterday as well, but didn't see it get through to the list and assume something must have gone wrong. Apologies in advance if it turns into a double post.