Örs Málnási-Csizmadia of Cogito Instruments SA

The highlight of that talk were a description of an innovative digital technology enabling embedded real-time learning and data analytics. This relies on Leeloo architecture fundations. The Multi chip neural network architecture is delivering significant advantages vs more traditional deep learning approaches:

• Real-time learning “on the job”
• Ability to learn from small datasets
• Protection of data privacy and model ownership
• Low power consumption
• Massive parallelism
• Flexible sensor data interface

Currently this technology is used in visual inspection, predictive maintenance and autonomous robotics applications with the available on the shelf module CI 9120.

Biography:

Email:

Address:Cogito Instruments SA, 2 rue de la Rôtisserie, Geneva, Switzerland, Switzerland, 1204 

Ning Qiao of aiCTX

Energy efficiency in computation is a major issue for enabling mobile/wearable, IoT and hyperscale computing applications. Von Neumann digital computation is reaching its limits in terms of fabrication feature size, and clock speeds are highly power-inefficient. A variety of methods are being developed to improve efficiency, including CPU clusters, GPUs, FPGAs and digital ASICs (accelerators). What these systems have in common is that they provide accelerated parallel computation for specialized tasks (e.g. machine vision), trading reduced clock speed for increased silicon area to achieve efficiency gains. Nevertheless, they are still bound by the inherent limitations of digital transistor physics.

aiCTX engineers spiking neural network and algorithmic solutions that implement computational neuroscience models of cortical computation. It focuses on developing end-to-end dedicated neuromorphic processors, and smart sensor SOC solutions for real-time ultra-low-power (sub-mW to mW) and ultra-low-latency latency (<30ms) sensory processing. The solutions are either based on ultra-low-power mixed-signal VLSI circuits, or on fully-asynchronous low-power digital designs (or both). The event-driven processors  and sensors can be used for applications that require ultra-low power (sub-mW to mW) always-on solutions, for IoT edge-based computing on mobile and embedded systems.

Biography:

Dr. Qiao is an expert ultra-low-power analog and mixed-signal/fully-asynchrnous neuromorphic engineer, and holds the position of Chief Executive Officer at aiCTX. Dr Qiao is leading designs of next-generation neuromorphic processors for real-time event-driven processing. He is the founder, CEO and CTO of aiCTX.

Email:

Address:Airgate Business Center, Thurgauerstrasse, 40, Zürich, Switzerland, Switzerland, 8050