Neural Networks in Hardware: Architectures, Products and Applications Clark S. Lindsey Guest Scientist Royal Institute of Technology Stockholm, Sweden lindsey@particle.kth.se www.particle.kth.se/~lindsey Original Date: 6-March-98 [August 26, 2002 - Update] These lecture pages are available in a zip file           Table of Contents 1. Overview  Why Hardware NNW Applications Amdahls Law Hardware vs Software Designer's Dilemma User's Dilemma Hardware NNW Apps -OCR -Speech Recognition -Touchpad -High EnergyPhysics     2. Hardware NNWs  Introduction a. Chips ---Ratings i.   Digital ----Designs ii.  Analog & Hybrids ----Designs iii. Neuromorphic b.  General vs Specialized c.   NNW Systems i.   Neurocomputers ii.  Accelerator cards d.  Other Approaches    3. Case Studies    C: Case Studies Adaptive Solutions CNAPS i.   Architecture ii.  Programming iii. ART-1 1. CN code 2. C code IBM ZISC i.    Architecture ii.   Recall iii. Learning Univ. of Trento TOTEM i.    Architecture ii.   RTS algorithm iii.  RTS Advantages iv.   RTS & TOTEM    4. Summary and Discussion  5. References and Links