Case of data without edges features

[Hana Jarraya]

Hi,

Can i process a database that the graphs are without edges features as input pystruct model?

[Andreas Mueller]

Hi.
Edge features are not necessary. There is the GraphCRF class.
Whether your task can be solved using pystruct I can not answer without details.

Andy

On 02/11/2016 11:27 AM, Hana Jarraya wrote:

Hi,

Can i process a database that the graphs are without edges features as input pystruct model?

--
You received this message because you are subscribed to the Google Groups "pystruct" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pystruct+u...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Hana Jarraya]

Hi,

Thank you for the answer.
Actually my data is dataset of graphs extracted from images of alphabetic Letters.
The case that those graphs are defined by nodes and edges and nodes attributes.
I used the GraphCRF model but the results of ssvm on those graphs is meaningless.
So i want to know more details about the inferences methods, i mean the mathematical formula or the references in a way I can understand how it works.

[Andy]

Hi.
What do you mean when you say "meaningless"?
The book on structured prediction in computer vision that I'm
referencing is pretty good, as are the papers by Joachims et al.
There is also an explanation in Murphy's book on machine learning, I think.

Andy

[Hana Jarraya]

Actually my goal is to understand the behaviour of the classifier with two classes.
The classification results that I got are not good. Accuracy is equal to 0.7.
And the non classified instances are quite similar to the well classified one.
So the problem here, that I am not using the correct model or I am choosing the wrong parameters, I don't know exactly.
That's why I am asking about more details of the algorithms of inference methods and learning parameters.
Or maybe you would give me more tips to better the model that I have to choose.