Project

EU H2020 NEURONN: Two-dimensional oscillatory neural networks for energy efficient neuromorphic computing

Goal: Neuro-inspired computing employs technologies that enable brain-inspired computing hardware for more efficient and adaptive intelligent systems. Mimicking the human brain and nervous system, these computing architectures are excellent candidates for solving complex and large-scale associative learning problems.

The EU-funded NeurONN project will showcase a novel and alternative neuromorphic computing paradigm based on energy-efficient devices and architectures. In the proposed neuro-inspired computing architecture, information will be encoded in the phase of coupled oscillating neurons or oscillatory neural networks (ONN).

The Consortium of six (6) partners is led by CNRS, the National Centre of Scientific Research (France). The project partners are IBM Research Zurich, Fraunhofer EMFT, CSIC/University of Seville, Silvaco, UK and AI Mergence, FR. Furthermore, NeurONN has initiated an Industrial Advisory Board which consists of members from Intel Corporation and Prophesee.

Updates
0 new
4
Recommendations
0 new
0
Followers
0 new
16
Reads
2 new
177

Project log

Aida Todri-Sanial
added 2 research items
In this work, we investigate by means of atomistic density functional theory simulations the interaction between cortisol (the target molecule) and monolayer MoS2 (the substrate). The aim is to assess viable strategies for the non-enzymatic chemical sensing of cortisol. Metal doping of the sensing material could offer a way to improve the device response upon analyte adsorption, and could also enable novel and alternative detection mechanisms. For such reasons, we explore metal doping of MoS2 with Ni, Pd, and Pt, as these are metal elements commonly used in experiments. Then, we study the material response from the structural, electronic, and charge-transfer points of view. Based on our results, we propose two possible sensing mechanisms and device architectures: (i) a field-effect transistor, and (ii) an electrochemical sensor. In the former, Ni-doped MoS2 would act as the FET channel, and the sensing mechanism involves the variation of the surface electrostatic charge upon the adsorption of cortisol. In the latter, MoS2 decorated with Pt nanoparticles could act as the working electrode, and the sensing mechanism would involve the reduction of cortisol. In addition, our findings may suggest the suitability of both doped and metal-doped MoS2 as sensing layers in an optical sensor.
The increasing amount of data to be processed on edge devices, such as cameras, has motivated Artificial Intelligence (AI) integration at the edge. Typical image processing methods performed at the edge, such as feature extraction or edge detection, use convolutional filters that are energy, computation, and memory hungry algorithms. But edge devices and cameras have scarce computational resources, bandwidth, and power and are limited due to privacy constraints to send data over to the cloud. Thus, there is a need to process image data at the edge. Over the years, this need has incited a lot of interest in implementing neuromorphic computing at the edge. Neuromorphic systems aim to emulate the biological neural functions to achieve energy-efficient computing. Recently, Oscillatory Neural Networks (ONN) present a novel brain-inspired computing approach by emulating brain oscillations to perform autoassociative memory types of applications. To speed up image edge detection and reduce its power consumption, we perform an in-depth investigation with ONNs. We propose a novel image processing method by using ONNs as a hetero-associative memory (HAM) for image edge detection. We simulate our ONN-HAM solution using first, a Matlab emulator, and then a fully digital ONN design. We show results on gray scale square evaluation maps, also on black and white and gray scale 28x28 MNIST images and finally on black and white 512x512 standard test images. We compare our solution with standard edge detection filters such as Sobel and Canny. Finally, using the fully digital design simulation results, we report on timing and resource characteristics, and evaluate its feasibility for real-time image processing applications. Our digital ONN-HAM solution can process images with up to 120x120 pixels (166 MHz system frequency) respecting real-time camera constraints. This work is the first to explore ONNs as hetero-associative memory for image processing applications.
Aida Todri-Sanial
added 3 research items
In this paper, we assess an innovative concept of emulating biological neurons with oscillators to implement an oscillatory neural network (ONN) with beyond-CMOS devices based on vanadium dioxide (VO2). ONNs can be of interest as an ultra-low-power neuromorphic architecture capable of performing associative memory tasks, such as pattern recognition in IoT edge devices. To explore the benefits and costs of beyond-CMOS ONNs necessitates modeling, simulation, and design methods spanning from materials (e.g., atomistic methods) to devices (e.g., technology-computer-aided-design, TCAD) up to circuits (e.g., mixed-mode simulation, compact modeling). In this work, we report on the development of such an advanced design toolbox and the results on performance and features of beyond-CMOS ONNs. The proposed design toolbox allows exploring ONN scalability, accuracy, energy, and performance for pattern recognition applications.
Aida Todri-Sanial
added 6 research items
Nano-oscillators based on phase-transition materials are being explored for the implementation of different non-conventional computing paradigms. In particular, vanadium dioxide (VO 2 ) devices are used to design autonomous non-linear oscillators from which oscillatory neural networks (ONNs) can be developed. In this work, we propose a new architecture for ONNs in which sub-harmonic injection locking (SHIL) is exploited to ensure that the phase information encoded in each neuron can only take two values. In this sense, the implementation of ONNs from neurons that inherently encode information with two-phase values has advantages in terms of robustness and tolerance to variability present in VO 2 devices. Unlike conventional interconnection schemes, in which the sign of the weights is coded in the value of the resistances, in our proposal the negative (positive) weights are coded using static inverting (non-inverting) logic at the output of the oscillator. The operation of the proposed architecture is shown for pattern recognition applications.
The collective behavior of many coupled oscillator systems is currently being explored for the implementation of different non-conventional computing paradigms. In particular, VO2 based nano-oscillators have been proposed to implement oscillatory neural networks that can serve as associative memories, useful in pattern recognition applications. Although the dynamics of a pair of coupled oscillators have already been extensively analyzed, in this paper, the topic is addressed more practically. Firstly, for the application mentioned above, each oscillator needs to be initialized in a given phase to represent the input pattern. We demonstrate the impact of this initialization mechanism on the final phase relationship of the oscillators. Secondly, such oscillatory networks are based on frequency synchronization, in which the impact of variability is critical. We carried out a comprehensive mathematical analysis of a pair of coupled oscillators taking into account both issues, which is a first step towards the design of the oscillatory neural networks for associative memory applications.
Aida Todri-Sanial
added an update
Aida Todri-Sanial
added an update
Jamila Boudaden
added an update
It is an exciting new research area that is becoming increasingly important in the context of the AI (artificial intelligence) megatrend: so-called neuromorphic computing uses technologies that imitate the human brain and nervous system. It is thus predestined to solve complex and comprehensive associative learning problems. At the same time, it offers the opportunity to significantly reduce the energy consumption of current silicon-based circuits.
In the EU project NeurONN, launched in early 2020, a research team from Fraunhofer EMFT is working with six European partners on a new neuromorphic approach based on energy-efficient elements and architectures. In the proposed neurologically inspired computer architecture, information is encrypted in the phase of coupled oscillating elements that are interconnected to form a neural network.  Just like the brain, the two key components in neuromorphic computing are called neuron and synapse - they replicate the distributed computing and memory units. The neurons used in the project are novel elements based on vanadium dioxide, which can be 250 times more efficient than state-of-the-art digital oscillators based on CMOS.
The work package of the Fraunhofer EMFT focuses particularly on the synapses: Within the framework of NeurONN, the scientists are developing 2D memristors on a nanoscale based on innovative 2D nanomaterials. The tiny devices are expected to be 330 times more efficient in terms of operating speed, lifetime and energy consumption than currently used technologies.
The project with a duration of 36 months (1 January 2020 - 31 December 2022) brings together leading European research and academic institutions: IBM Research Zurich, the Fraunhofer EMFT, CSIC/University of Seville, Silvaco, UK and AI Mergence, FR. It is coordinated by the French CENTRE NATIONAL DE LA RECHERCHE SCIENTIFIQUE CNRS. Additionally, NeurONN has initiated an industrial advisory board including members from Intel Corporation and Prophesee.
The NeurONN kick-off meeting took place in Montpellier (France) on February 4 and 5, 2020 at the premises of LIRMM, CNRS. The project is funded under the EU research program Horizon 2020 under grant number 871501
Coordinator Contact info:
Dr. Aida Todri-Sanial Project Coordinator CNRS Director of Research
 
Jamila Boudaden
added an update
It is an exciting new research area that is becoming increasingly important in the context of the AI (artificial intelligence) megatrend: so-called neuromorphic computing uses technologies that imitate the human brain and nervous system. It is thus predestined to solve complex and comprehensive associative learning problems. At the same time, it offers the opportunity to significantly reduce the energy consumption of current silicon-based circuits.
In the EU project NeurONN, launched in early 2020, a research team from Fraunhofer EMFT is working with six European partners on a new neuromorphic approach based on energy-efficient elements and architectures. In the proposed neurologically inspired computer architecture, information is encrypted in the phase of coupled oscillating elements that are interconnected to form a neural network. Just like the brain, the two key components in neuromorphic computing are called neuron and synapse - they replicate the distributed computing and memory units. The neurons used in the project are novel elements based on vanadium dioxide, which can be 250 times more efficient than state-of-the-art digital oscillators based on CMOS.
The work package of the Fraunhofer EMFT focuses particularly on the synapses: Within the framework of NeurONN, the scientists are developing 2D memristors on a nanoscale based on innovative 2D nanomaterials. The tiny devices are expected to be 330 times more efficient in terms of operating speed, lifetime and energy consumption than currently used technologies.
The project with a duration of 36 months (1 January 2020 - 31 December 2022) brings together leading European research and academic institutions: IBM Research Zurich, the Fraunhofer EMFT, CSIC/University of Seville, Silvaco, UK and AI Mergence, FR. It is coordinated by the French CENTRE NATIONAL DE LA RECHERCHE SCIENTIFIQUE CNRS. Additionally, NeurONN has initiated an industrial advisory board including members from Intel Corporation and Prophesee.
The NeurONN kick-off meeting took place in Montpellier (France) on February 4 and 5, 2020 at the premises of LIRMM, CNRS. The project is funded under the EU research program Horizon 2020 under grant number 871501.
 
Aida Todri-Sanial
added a project goal
Neuro-inspired computing employs technologies that enable brain-inspired computing hardware for more efficient and adaptive intelligent systems. Mimicking the human brain and nervous system, these computing architectures are excellent candidates for solving complex and large-scale associative learning problems.
The EU-funded NeurONN project will showcase a novel and alternative neuromorphic computing paradigm based on energy-efficient devices and architectures. In the proposed neuro-inspired computing architecture, information will be encoded in the phase of coupled oscillating neurons or oscillatory neural networks (ONN).
The Consortium of six (6) partners is led by CNRS, the National Centre of Scientific Research (France). The project partners are IBM Research Zurich, Fraunhofer EMFT, CSIC/University of Seville, Silvaco, UK and AI Mergence, FR. Furthermore, NeurONN has initiated an Industrial Advisory Board which consists of members from Intel Corporation and Prophesee.