NeuroMem for Real-time Machine Learning

3 gears to try trainable, responsible and explainable AI Interested in Training?Advanced Tools

Gear #1: NeuroMem simulation

  • 1024 neurons
  • Von Neumann limitation
    • No deterministic latencies

Gear #2: NeuroMem on FPGA

  • 128 neurons (custom capacity on demand)
  • Parallel architecture
    • Witness deterministic latencies

Gear #3: NeuroMem chip


Open source simulation

  • C++ source code
  • C++ behavioral test script
    • Visual Studio project
    • Xilinx SDK project with ZYNQ bsp

IP core on FPGA

  • NGO file for Xilinx ZYNQ 7020
  • Verilog behavioral test script
  • Tested on Arty, MiniZed and SnickerDoodle

NeuroMem chip

  • NeuroShield board
  • SDK for Arduino, Python and C++
  • HDK for Xilinx ZYNQ boards
  • C++ behavioral test script