Select Language

SoC White Box IPs

AI-Neural Network Scalable Processor Core IP

OverviewFeaturesRequest Datasheet

Description

Spiking neural networks implemented with SNAP core operate in parallel, making them significantly faster than software neural networks running on Central Processing Units (CPUs) or Graphics Processing Units (GPUs). SNAP is more energy efficient, enabling SNN to be integrated into portable devices for local processing of sensor data. SNAP based neural networks can respond in real time with low latency, regardless of the neural network size. SNAP implements learning rules in hardware, enabling Autonomous Features Extraction (AFE) directly from input data without need for any software processing .
SNAP based neuromorphic chips can be used in nearly any embedded systems that requires pattern recognition or AFE locally without having to go to cloud based computations. Such systems represent a massive and unlimited potential market, with applications in Smartphones, Internet of Things (IoT), machine to machine (M2M), robotics, gaming, driverless vehicles, drones and air transportation among others.
snap

Features

  • Neural Network updates at 1MHz
  • Fully parallel computation
  • Real time pattern recognition
  • Input from DVS camera
  • Unsupervised feature learning
  • Spike Timing Dependent Plasticity
  • Library of learned features
  • Labeled output (vehicles counts)
  • Benefits

  • Fast, independent of size
  • Highly energy efficient
  • Real time response
  • Unsupervised Learning
  • Actionable data
  • Customizable Network
  • Applications

  • Surveillance and security cameras
  • Smartphones
  • Internet of Things (IoT)
  • Machine to machine (M2M)
  • Robotics, Gaming
  • Driverless Vehicles
  • Drones and Air Transportation
  • Deliverables

  • SNAP based Spiking Neural Network RTLv
  • Configurable Neuron and Synapses
  • Configurable Learning rules
  • Configurable synaptic connectivity
  • Proprietary spike communication protocol
  • Direct connection to sensory inputs
  • Creation of libraries of learned behaviors
  • SNAP Models Integrated with Neural Network Simulators like NENGO*
  • Design services for SNAP customization
  • Hardware emulation in FPGA
  • Configuration management tools

Please fill in the form below

T2M's range of high quality pre-verified, analog/mixed-signal, RF, Digital and SW system solutions, are used as critical building blocks of communications, consumer and computer products including IoT, Wearables, cellular, tablet, M2M, RCU, set-top boxes, TVs, DVD players and PC chipsets. IPs can be modified to meet the customer's specific requirement be it fab/node porting or proprietary features.