GANN Neural Network

[Khokhaz]

To who it may concern, 

I am a student in SAS Data Science Academy.

Currently studying

Module 4: Advanced Predictive Modeling :

Neural Network Modeling : 

Chapter 4: The NEURAL Procedure Generalized Additive Neural Networks

I have a question regarding GANN : is there a simple formula to calculate number of parameter estimates given the number of inputs and number of hidden unit per each input. ? 

Appreciate your assistance supported by examples. 

Thanks & Regards, 

[Cynthia_sas]

Hi:
  One of the instructors replied with this information:

=== ===

I’m interpreting GANN as an ordinary, fully connected feed forward NN; each input is in each hidden unit (node), and each hidden unit has a bias, so it’s (inputs + 1)*nodes. There are also weights in the linear combination of nodes that also has an intercept (bias) so add #nodes + 1 to above.

=== ===

  Another instructor asked for more clarification:

=== ===

  Is the student referring to the Generalized Additive Neural Networks (GANN) that we discuss in the self-study section in the course, as shown below:

  Is this the GANN he's asking about? Or is he referring to another type of GAN, which is known in the literature as generative adversarial network? If he's referring to the Generalized Additive Neural Network in the self-study section, can he provide a page number or example in the self-study PDF that he's asking for clarification about?

=== ===

  If your answer is supplied by the self-study material or by the first instructor's answer, then there's no need to reply further. However, if you have more clarification of your question, we can ask the instructors for more in-depth feedback. Please be aware that SAS will be closed for Thanksgiving on 11/28 and 11/29 and will re-open on Dec 2. 

Cynthia