7
\$\begingroup\$

I am doing a thesis project. I need to make a FPGA implementation of a neural network. I want to implement some mathematical functions such as an exponential function.

Does anyone know of any good books about FPGAs, VHDL or digital architecture implementation of floating point arithmetic operations?

\$\endgroup\$
4
  • 1
    \$\begingroup\$ What's your skill level? Do you need an intro to FPGAs and VHDL, or a reference for implementing floating-point operations with some recipes in VHDL? \$\endgroup\$ Jan 13, 2011 at 23:06
  • \$\begingroup\$ @reemrevnivek: I´ve studied VHDL before, but I haven´t had a formal intro to FPGAs.Do you know any reference I should read about FPGAs? I have the feeling FPGAs is a wide world in the sense, every FPGA manufacturer has a different technology. Of course theres is something in common between all manufacturers... \$\endgroup\$
    – Peterstone
    Jan 14, 2011 at 11:34
  • \$\begingroup\$ Do you have a project or class page link? This sounds very interesting. \$\endgroup\$
    – tyblu
    Jan 16, 2011 at 18:23
  • \$\begingroup\$ this has been converted to community wiki by your request. \$\endgroup\$
    – Kortuk
    Jan 29, 2011 at 15:50

4 Answers 4

8
\$\begingroup\$

There are ANN techniques that don't require floating-point. You will find things much easier if you adopt one of those.

Jim Austin's work at York University comes to mind. He uses FPGAs and fixed-point DSPs.

EDIT: (Leon, I hope you don't mind me adding this list here)

\$\endgroup\$
5
  • \$\begingroup\$ It´s difficult to me imagine something out of ecuations or whatever mathematical expressions. Well, at the end you have mathematical expressions and how do you implement it? That´s the questions.I suposse neural networks are interesting because they can be implemented on electronics (I don´t know imagine any signal processing application that it not based on mathematical expressions). Anyway, I have to use a Fuzzy Neural Networks and as I am concerned, Fuzzy Neural Networks are described by mathematical expressions. \$\endgroup\$
    – Peterstone
    Jan 13, 2011 at 15:25
  • \$\begingroup\$ Besides, I use the 'if then rule' form proposed by Takagi and Sugeno. This form is described by mathematical expressions. \$\endgroup\$
    – Peterstone
    Jan 13, 2011 at 15:27
  • \$\begingroup\$ Please, can you show me any reference about ANN techniques that don't require floating-point? \$\endgroup\$
    – Peterstone
    Jan 13, 2011 at 15:28
  • \$\begingroup\$ The link you have put it isn´t usefull, that web page don´t contain any link to documentation or libraries. May be I can access if I were a member of York University but I am not. \$\endgroup\$
    – Peterstone
    Jan 15, 2011 at 8:26
  • \$\begingroup\$ @Peterstone: it contains links to all the publications by that research group. That is the normal way for scientists to share their knowledge. \$\endgroup\$
    – drxzcl
    Feb 16, 2011 at 9:20
5
\$\begingroup\$

Professor Miriam Leeser's research group at the Reconfigurable Computing Laboratory at Northeastern University has developed an open source floating point library for Xilinx FPGAs in VHDL: http://www.ece.neu.edu/groups/rcl/projects/floatingpoint/index.html

\$\endgroup\$
4
\$\begingroup\$

They are starting on a package at VHDL.org. OptNgn also have a library that is free (GPLv3).

\$\endgroup\$
1
  • \$\begingroup\$ Both links are dead now. \$\endgroup\$
    – Peter K.
    Oct 17, 2018 at 15:26
2
\$\begingroup\$

I would avoid IEEE floating point format, go with the format the TI DSP uses. Ti has some good doc that goes along with one of the older DSPs that holds your hand through the operations. From there you can start on your logic.

\$\endgroup\$
2

Not the answer you're looking for? Browse other questions tagged or ask your own question.