[comp.ai.neural-nets] Backprop upgrades reference list

mgj@cup.portal.com (Mark Gregory Jurik) (04/23/91)

               CALL FOR YOUR FAVORITE REFERENCES ON  BACKPROP UPGRADES

I am compiling a list of references on all respectable improvements to BackPackProp.
This list will then be posted.

The following topics are included:

    1. The effect of low bit quantization
    2. The effects of momentum and batching (experimental proof needed)
    3. QuickProp, FastProp, UniProp,....whateverProp
    4. Conjugate Gradients
    5. 2,3 or more hidden layers
    6. Added noise in the data 
    7. Added noise in the weight adjustments
    8. Gaussian hidden layer cells  
    9. Kohonen layer pre-processing of the raw data
   10. Pre-processing to remove (and later add back) linear transformations
   11. Pre-processing to orthogonalize the input data

If you have any other topics that ought to be added to this list, please let
me know.

If you have favorite references addressing any of these topics, please let me 
know.

Please ask authors to send their most definitive papers on any of the above
topics to the following address.  I am co-authoring a book with Philip 
Wasserman and wish to include summaries of as much of the above topics as is
possible.  Thanks.

                 Jurik Research & Consulting
                 PO 2379
                 Aptos, CA   95001    USA

----------------------------------------------------------------------------