Wiki

Clone wiki

Artificial Neural Networks / Home / FPGAsAndNeuralNets

FPGAs and Neural Networks

This page contains useful resources for building Neural networks on FPGAs.

This also has FPGA boards that we can use for trying this out.

SBCs and IoT Boards

  1. Intel discontinues Joule, Galileo and Edison product lines

FPGA Development Boards

FPGA vendors are mainly Altera (bought by Intel now) and Xilinx. (TODO: Need to find the other top vendors of FPGAs). Given below are links to these product documentations and also other comparison articles.

  1. Altera's Cyclone Series of FPGAs, they do not recommend Cyclone II and older chips for new designs.
  2. Altera's Cyclone II EP2C5T144 Minimum Sys DevBoard available in eBay India for INR.1874
  3. TODO: Find out the cheapest Altera Cyclone IV or V Dev board ??
  4. SnickerDoodle: Zynq 7000 Series FPGA Board with Processor, available from Crowdsupply, 1.3M ASIC gates, 180 reconfig. I/O, 1GB LPDDR2 RAM, with SDSoC license: USD.195

FPGAs for Neural Networks

Use of FPGAs and ASICs for building Neural Networks, is one of the recent developments. Some references on this is listed below:

Taken from Ref #2 below:

DeePhi is a relatively new company (launched March 2016) based on efforts 
from teams at Stanford and Tsinghua University, As the startup’s CEO and co-founder, Song Yao, described at Hot Chips this week in his introduction of DeePhi (which is short for the phrase “discovering the philosophy behind deep learning computing”),
the economics and time to market pressures matched with the rapid evolution of deep learning frameworks make non-FPGA approaches more expensive and less efficient.

Yao says that CPUs don’t have the energy efficiency, GPUs are great for training but lack “the efficiency in inference”, DSP approaches don’t have high enough performance and have a high cache miss rate and of course, ASICs are too slow to market—and even when produced, finding a large enough market to justify development cost is difficult.
  1. FPGA Aceleration of Convolutional Neural Network, NallaTech Whitepaper
  2. FPGA Based Deep Learning Accelerators Take on ASICs
  3. Why are GPUs necessary for training deep learning models?, AnalyticsVidhya, May 2017
  4. TODO: Add other links for Google search results: '"deep learning" processing unit'

FPGAs and VHDL COurses

  1. Master VHDL Design for use in FPGA and VLSI Digital Systems Updated Nov 2016
  2. TODO: Check which course I bought long time ago & buy any new & latest course.

Updated