hi Guys, I am doing a project in nueral nets and I wish to program it as opposed to finding code (usually messy and uncommented) on the net and tweaking it. But the only thing I find hard is learning it. I understand feed-forward NNs but dont understand how I would implement backpropagation. Every time I read about backprop all literature confuse me with so much maths. I know the maths is important but they never give practical examples of backprop and SOMs they just give you the mathematics with complex symbols and when they list the algorithms they throw in all that mathematics to confuse you. It would be better if they explained it in plain English and then gave maths to support it. Not the English to support the maths. Does anyone know if they have ever released a book that explains NNs soooo dumbed down. Like the "...For Dummies" series of books. They always take complex subjects and dumb it down while still teaching you. Or any web sites with the most simplest descriptions of NNs which dumbs it down while still gets the point across. Im not afraid of the maths I just cannot stand explainations where the maths is the only thing thrown at you and no much theory. Its funny I was reading a book the other day called Introduction to Neural Networks and on the back it said it gives a comprehensive guide to NNs without confusing the reader but when I opened it the content was very confusing. Can anyone suggest. I have only a few weeks left before I commence this final year project. Thanks. DarkStar UK ------------------------------- Loves cross-posting because it works

I think my book''s exactly what you''re looking for...



My Website: ai-junkie.com | My Book: AI Techniques for Game Programming

Oh Ive been on to ur site fup,(ai-junkie) I like ur tutorials. So ur book uses more explainations than maths right? (not that I hate maths) but most are sooo overwhelming with the maths part. I need something to really explain backprop... so clear that even without much maths or code I can do one myself. Because after reading a tutorial it was so clear I implemented feed-forward NN and this was without code or much maths.

Thanks

DarkStar
UK

My book has a chapter on how to program and use backpropagation. I use a practical example and demonstrate how the technique can be improved. I have deliberately *not* gone into the math. I show the reader the equations but do not go into their derivation. I’ve tried to make it as easy to follow as possible.



My Website: ai-junkie.com | My Book: AI Techniques for Game Programming

Ok then, ive had a look at the book on Amazon and on this site. I like the reviews. Im thinking of purchasing it now. I need as much help as possible with this topic. Thanks for the help.

Thanks

DarkStar
UK

-------------------------------
Loves cross-posting because it works

quote:

---

Original post by Dark Star
I am doing a project in nueral nets and I wish to program it as opposed to finding code (usually messy and uncommented) on the net and tweaking it. But the only thing I find hard is learning it. I understand feed-forward NNs but dont understand how I would implement backpropagation. Every time I read about backprop all literature confuse me with so much maths.

I know the maths is important but they never give practical examples of backprop and SOMs they just give you the mathematics with complex symbols and when they list the algorithms they throw in all that mathematics to confuse you. It would be better if they explained it in plain English and then gave maths to support it. Not the English to support the maths.

---

First, let me offer my sympathy. Backpropagation is an important tool, but I, too, have found most published explanations of its operation wanting. If you are sure that backpropagation is what you want, I suggest "Neural Networks for Statistical Modeling" by Murray Smith.

You might also consider any of the dozens of other neural architectures. Many of them are much simpler conceptually than backpropagation-trained FFNNs. One of the easiest to understand is RCE ("Restricted Coulomb Energy"). These are a bit scattered, but this was all I could find in 5 minutes:

http://www.fee.vutbr.cz/~grebenic/Publikace/Mosis99/
http://www.warthman.com/images/Ni1000c2.pdf
http://www.dontveter.com/basisofai/11.html

good luck,
Predictor
http://will.dwinnell.com