![]() One of Five Cyclone Boards |
• Bill led his Motorola team’s creation of a 2000 Processor System, built from 500 “quad-neuron” chips (full custom, 100% digital, CMOS integrated circuits). • The massively parallel array of neuron-like processors executed fast self-organizing maps, using an efficiently pipelined “ring” architecture. • Keywords: Hardware, CMOS, Massively Parallel, Pipelined, Self-Organizing Maps, Formal Neurons, UNIX. * Cyclone (a.k.a. Cyclonne) comprised 500 “Fan” (a.k.a. Fann) chips. |
![]() |
• Cyclone was used to accelerate Motorola R&D such as the team’s highly successful CorDex product. • Cyclonne allowed us to tackle large databases (e.g. stock market). • Cyclonne demonstrated that the “computationally intensive element” (Self-Organizing Maps a.k.a.SOMs) of our neuroscience-inspired architectures could be made extremely fast, and power-efficient. |
![]() |
Leadership / Management • Bill led the entire Cyclonne project. (Mike Gardner, a member of Bill’s team, was an outstanding contributor, too). Theory / R&D • Bill focused his team on SOMs, despite industry-wide pressures to focus on Backpropagation. • This includes Bill’s numerous insights in machine learning, neuroscience, computer science, cognitive science and hardware design (CMOS design, pipelining, memory architectures, ALUs and microcode. Software • Bill led the team’s system/application integration of Cyclone into UNIX-based VME platform. (Jack Bieker, a member of Bill’s team, contributed significantly to this task). • All Cyclonne applications employed significant software. Hardware • First silicon was successful. Bill’s effective design reviews helped accomplish this. Bill’s experience in logic, memory, microcoded ALUs, and silicon design kept the design on track. |
![]() |
Patent (5, 216, 751) Issued (6 - 1 -1993) Current Owner: Motorola, Inc. Licensing availability: (Please contact Motorola). Contact Info: motorolaventures@motorola.com |
![]() |
By pushing beyond the current state-of-the-art, there are these opportunities: • The human brain (neocortex) contains many hundreds of self-organized maps and there are numerous valuable humans skills that still defy computer implementation. • We can learn from nature’s brilliance, or we can ignore it. Advanced SOM-like hardware learns from it. • Today, Intel and AMD multiprocessors beginning to sell in volume. Yet, software that efficiently uses them is lacking. Cyclonne-like systems have obvious parallelism that can very efficiently use multiprocessors to execute valuable human-like applications. |
![]() |
• Patent 5216751 |
