Non-parametric model to graph classification

[gkiar]

In class today, we talked about different graph models and discussed the pros and cons based on the number of parameters (extreme cases being 1 parameter and 2^n^2 parameters). What are some non-parametric models for graph inference/classification, and what are their pros/cons?

[jovo]

the categorical model is non-parametric. do you know why?

my signal subgraph model from my PAMI paper is semi-parametric. do you know why?

On Tuesday, February 3, 2015, Greg Kiar notifications@github.com wrote:

In class today, we talked about different graph models and discussed the pros and cons based on the number of parameters (extreme cases being 1 parameter and 2^n^2 parameters). What are some non-parametric models for graph inference/classification, and what are their pros/cons?

— Reply to this email directly or view it on GitHub https://github.com/Statistical-Connectomics-Sp15/intro/issues/44.

the glass is all full: half water, half air. openconnecto.me, we're hiring! https://docs.google.com/document/d/14SApYAzxF0Ddqg2ZCEwjmz3ht2TDhDmxyZI2ZP82_0U/edit?usp=sharing , jovo.me, my calendar https://www.google.com/calendar/embed?src=joshuav%40gmail.com&ctz=America/New_York

[mblohr]

A categorical model is nonparametric because it assumes no knowledge about the form of the data. The signal subgraph model is semiparametric because the joint distribution over all graphs and classes is parametric (at first glance it looks nonparametric because of the indexing, but these indices are actually a function of model parameters determined as a first step in this algorithm) and the classifier is a nonparametric argmax, over an index of classes, of the joint distributions calculated in the first step.