NEUROMEM® CHIPS

Trainable

Learn on the chip. Stimuli can derive from text, bio-signals, audio files, images and videos, etc.

Real Time

Recognize patterns in micro-seconds regardless of the number of models stored in the neurons.

Energy Efficient

>100K recognitions per second at 27Mhz. Less than 300 milliWatts per 1000 neurons.

Available NOW

Available on ASICs, SOCs, and FPGAs and can be evaluated on a variety of platforms

Digital neurons inspired by biology

  • All interconnected and working in parallel following a Winner-Takes-All collective behavior
  • Recognize and learn a pattern in a constant time regardless of the number of neurons committed in the chip
  • Behave collectively as  a KNN (K-Nearest Neighbor) or RBF classifier (Radial Basis Function).
  • Cope with ill-defined and fuzzy data, high variability of context and novelty detection.
  • Multiple NeuroMem chips can be daisy-chained to scale a network from 1000s to millions of neurons with the same speed and simplicity of operation as a single chip.
  • Amazingly simple to deploy

The Flavors of NeuroMem ICs…

 

Today, several flavors of NeuroMem chips are available for your designs, but regardless of the chip you select, the access to the neurons will remain identical and their knowledge will be portable accross platform. This is made possible because all NeuroMem chips use the same bus and register map. Read more>>.

 

CM1K

  • 1024 neurons
  • 256 bytes of memory per neuron
  • Cascadable
  • I2C (optional use)
  • Recognition stage (optional use)

NM500 chip

  • 576 neurons
  • 256 bytes of memory per neuron
  • Cascadable
  • Chip Scale Package

NeuroMem IP

  • NeuroMem IP for FPGA or SOC
  • Custom neuron capacity
  • Custom memory capacity per neuron

Intel Quark SE

  • 128 neurons
  • 128 bytes of memory per neuron
  • Component of an SOC also combining an x86 MCU and a sensor subsystem
gv