search

open-connectome-classes / StatConn-Spring-2015-Info

introductory material
18 stars 4 forks source link

Exponential Families? #135

Open ElanHR opened 9 years ago

ElanHR commented 9 years ago

In many papers the authors fall back on their model being part of an exponential family to justify the use of a number of nice properties (in this paper it is in the "exponential family of random graph models").

Are there any good resources to understand what is meant by this and what this gives us?

yaxigeigei commented 9 years ago

Please check this online course. 5.1 - 5.2 introduce what is the exponential family. 7.4 shows the significance in choosing prior for tractable calculation. From 7.5 - 7.6, you will see the prior distribution of theta in Bern(theta), such as those in SBM, can be expressed as Beta(a,b), which is the distribution of our mean connectome! https://www.youtube.com/playlist?list=PLD0F06AA0D2E8FFBA

ajulian3 commented 9 years ago

When using exponential families of data in neural networks, how does one stabilize the data? I understand this is a new development of neural research.