0

0

I came across a paper where they are predicting GPA/Grades of different courses for a candidate using RBM.

My question is can we apply RBM for a problem where the output needs to be continuous and multiple?? like a students' score on different courses or case size of different product that a customer can buy...etc

If yes then a study material / link that help me start with will be helpful!!

Thanks in advance.

S.alam
  • 1

1 Answers1

0

This is theoretically possible to build an energy with both real inputs and outputs. Hinton's technical report gives a word about the Gaussian-Gaussian RBM in Section 13.3. However he also warns that in practice this model is very unstable. I have personally no experience with them.

Provided you furthermore want multinomial units in your RBM (which is done in this article for example), you might end up with a rather impractical model.

TheCG
  • 822
  • 4
  • 13
  • Thanks @TheCG .. I think then RBM is not a good choice in predicting GPA/Score prediction for different courses for a candidate and/or alike problems? However, any suggestions regarding what techniques could be explored regarding the problem at hand would be really helpful? – S.alam May 23 '19 at 05:29
  • Maybe you could try anyway such a network with gaussian-gaussian units! This [paper](http://www.iro.umontreal.ca/~lisa/pointeurs/BengioNips2006All.pdf) seems to explore it into more details! – TheCG May 23 '19 at 06:18
  • Moreover, after a google search, i found this document [https://arxiv.org/pdf/1708.08744.pdf](https://arxiv.org/pdf/1708.08744.pdf) that deals with your problem and proposes to solve with a RBM! – TheCG May 23 '19 at 06:19
  • Thansk a lot @TheCG, actually I started with this paper only...but the RBM solution for this is not detailed enough for the problem I have at hand! But I think now I have some directions to approach a solution framework...helpful! – S.alam May 25 '19 at 05:53