It is a really nice combo! I got clued in that the variable was a 2D
array, and once I saw that the rest was easy. :) Now I just have to walk
the feature weights back and see how significant these things are! After
double checking that the arrays are of the same size, of course. :)
Thanks Greg!
Matthew
On Thu, Feb 19, 2015 at 10:29 PM, Greg Landrum <[email protected]>
wrote:
>
>
> On Thu, Feb 19, 2015 at 11:59 PM, Matthew Lardy <[email protected]> wrote:
>
>>
>> I have been able to build models via scikit-learn with the RDKit python
>> wrappers. That all works beautifully!
>>
>
> It's a nice combination, isn't it?
>
>
>> What I am struggling to get are the weights, or scalers, applied to each
>> bit position. For a SVM regression model (SVR) I think that the values I
>> seek are in the coef_ (if the model is created via the linear kernel).
>> But, all I get is something like this when I print that out:
>>
>> [[-0. -0.87146158 -0.46331996 ..., 0.31076767 -0.
>> -0.81882195]]
>>
>>
> I don't really know the SVM regression approach particularly well, but it
> looks like that's a vector of vectors. Is the length of the inner vector
> the same as the length of the fingerprint/descriptor vector you are
> providing?
>
> -greg
>
>
------------------------------------------------------------------------------
Download BIRT iHub F-Type - The Free Enterprise-Grade BIRT Server
from Actuate! Instantly Supercharge Your Business Reports and Dashboards
with Interactivity, Sharing, Native Excel Exports, App Integration & more
Get technology previously reserved for billion-dollar corporations, FREE
http://pubads.g.doubleclick.net/gampad/clk?id=190641631&iu=/4140/ostg.clktrk
_______________________________________________
Rdkit-discuss mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/rdkit-discuss
