1. Biologically Inspired Computation Chris Diorio Computer Science & Engineering University of Washington diorio@cs.washington.edu

2. C. Diorio, 10–8–002 Nature is telling us something... F Can add numbers together in nanoseconds íHopelessly beyond the capabilities of brains F Can understand speech trivially íFar ahead of digital computers í…and Moore’s law will end

3. C. Diorio, 10–8–003 Problem: How do we build circuits that learn F One approach: Emulate neurobiology íDense arrays of synapses

4. C. Diorio, 10–8–004 Silicon synapses F Use the silicon physics itself for learning íLocal, parallel adaptation íNonvolatile memory

5. C. Diorio, 10–8–005 Silicon synapses can mimic biology F Local, autonomous learning

6. C. Diorio, 10–8–006 Synaptic circuits can learn complex functions F Synapse-based circuit operates on probability distributions íCompetitive learning íNonvolatile memory í11 transistors í0.35µm CMOS íSilicon physics learns “naturally” F Silicon learning circuit versus software neural network íBoth unmix a mixture of Gaussians íSilicon circuit consumes nanowatts â Scaleable to many inputs and dimensions

7. C. Diorio, 10–8–007 Technology spinoff: Adaptive filters F Synapse transistors for signal processing í~100× lower power and ~10× smaller size than digital Mixed-signal FIR filter 16-tap, 7-bits 225MHz, 2.5mW Built and tested in 0.35µm CMOS Adjust synaptic tap weights off-line FIR filter with on-chip learning 64 taps, 10 bits, 200MHz, 25mW In fabrication in 0.35µm CMOS On-line synapse-based LMS

8. C. Diorio, 10–8–008 Problem: How to study neural basis of behavior F Measure neural signaling in intact animals íImplant a microcontroller in Tritonia brain F Tritonia is a model organism íWell studied neurophysiology í500µm neurons; tolerant immune response íWork-in-progress

9. C. Diorio, 10–8–009 An in-flight data recorder for insects F An autonomous microcontroller “in-the-loop” íStudy neural basis of flight control Manduca Sexta or “hawk moth”