Periodic Reporting for period 3 - PHRESCO (PHotonic REServoir COmputing)
Berichtszeitraum: 2018-11-01 bis 2019-10-31
The PHRESCO project focuses on the development of efficient cognitive computing using photonics silicon-based technology. The final objective of PHRESCO is to design a new reservoir computing chip, including innovative electronic and photonic components allowing major advances in the field. The consortium of the project is composed by the Katholieke Universiteit Leuven (KUL) in Belgium, who is the coordinator, the Universiteit Gent (UGent) in Belgium, the Leibniz Institute for Innovative Microelectronics (IHP) in Germany, IBM Research GmbH in Switzerland and the Centrale Supélec (CS) in France.
New computing paradigms are required to feed the next revolution in Information and Communication Technology (ICT). Machines that can learn, but also handle vast amount of data, need to be invented. In order to achieve this goal and still reduce the energy footprint of ICT, fundamental hardware innovations must be done. Most of the time, CMOS is used to emulate e.g. neuronal behavior, but is intrinsically limited in power efficiency and speed. On the other hand, Reservoir Computing (RC) is one of the new computing methods that has proven its efficiency to perform tasks where traditional approaches fail. It is also one of the rare concepts of an efficient hardware realization of cognitive computing into a specific, silicon-based technology.
The overall objectives of the projects are i) to scale optical RC systems ii) build an all-optical chip based on the unique electro-optical properties of new materials iii) implement new learning algorithms to exploit the capabilities of the RC chip.
The main conclusions of the action are the folowing:
- Demonstration of upscaling of the optical reservoir system
- Integration of functional new materials with PIC technology
- Development of new training methods of the reservoir
New computing paradigms are required to feed the next revolution in Information and Communication Technology (ICT). Machines that can learn, but also handle vast amount of data, need to be invented. In order to achieve this goal and still reduce the energy footprint of ICT, fundamental hardware innovations must be done. Most of the time, CMOS is used to emulate e.g. neuronal behavior, but is intrinsically limited in power efficiency and speed. On the other hand, Reservoir Computing (RC) is one of the new computing methods that has proven its efficiency to perform tasks where traditional approaches fail. It is also one of the rare concepts of an efficient hardware realization of cognitive computing into a specific, silicon-based technology.
The overall objectives of the projects are i) to scale optical RC systems ii) build an all-optical chip based on the unique electro-optical properties of new materials iii) implement new learning algorithms to exploit the capabilities of the RC chip.
The main conclusions of the action are the folowing:
- Demonstration of upscaling of the optical reservoir system
- Integration of functional new materials with PIC technology
- Development of new training methods of the reservoir
The main outcomes of PHRESCO can be considered as significant advances in the following main areas: integration of ferroelectric technology for novel modulators and switches and analog signal processing for high speed datacom/telecom applications. In the first one, the work done during the project shows then need to continue improving the reliability and yield that can be achieved together with the demonstration of relevant high-speed devices. At the same time, it is important to identify risk-taking stakeholders in order to bring the developed hardware components to the maturity level necessary for real world applications.
The new ferroelectric technology developed through the course of PHRESCO has shown to have great interest in the photonics community due to its wide range of applications. Although great advances regarding the integration and stand-alone functionalities of both BaTiO3 weighting elements and III/V based amplifiers have been achieved, it was not possible to obtain a functioning demonstrator with both materials integrated in a large scale silicon photonic reservoir. Nevertheless, two important results of the PHRESCO project are the co-integration of BaTiO3 and III-V materials on a photonic platform and the experimental demonstration of non-linear dispersion compensation in an optical fiber line using a 1550nm reservoir with 32 nodes at record speed of 32 Gbps.
Other relevant developments related to system architecture and training methods are worth to mention. For example, novel architectures using non-linear components as nodes have been numerically studied showing that they can perform at state-of-the-art level on various benchmark tasks. On the other hand, new training methods that can deal with limited state observability as well as limited accuracy of the weighting elements have been developed. Remarkably, this explorative retraining method with only 16 levels for the weighting elements shows a performance close to a full-precision method. Finally, a detailed study of different possible hierarchical architectures for multi-reservoir systems has been done, concluding that ensembling and chaining are the best options.
The new ferroelectric technology developed through the course of PHRESCO has shown to have great interest in the photonics community due to its wide range of applications. Although great advances regarding the integration and stand-alone functionalities of both BaTiO3 weighting elements and III/V based amplifiers have been achieved, it was not possible to obtain a functioning demonstrator with both materials integrated in a large scale silicon photonic reservoir. Nevertheless, two important results of the PHRESCO project are the co-integration of BaTiO3 and III-V materials on a photonic platform and the experimental demonstration of non-linear dispersion compensation in an optical fiber line using a 1550nm reservoir with 32 nodes at record speed of 32 Gbps.
Other relevant developments related to system architecture and training methods are worth to mention. For example, novel architectures using non-linear components as nodes have been numerically studied showing that they can perform at state-of-the-art level on various benchmark tasks. On the other hand, new training methods that can deal with limited state observability as well as limited accuracy of the weighting elements have been developed. Remarkably, this explorative retraining method with only 16 levels for the weighting elements shows a performance close to a full-precision method. Finally, a detailed study of different possible hierarchical architectures for multi-reservoir systems has been done, concluding that ensembling and chaining are the best options.
The technologies that has matured the most during the first phase of the project, namely the photonic reservoir technology and the research focusing on barium titanate, will be described respectively focusing on their progression beyond the current state-of the-art.
On the level of photonic reservoir computing, the first generation prototype of Phresco goes beyond the state-of-the-art, since it will be the first realisation of a passive photonic reservoir with integrated all-optical readout. This will allow for high-speed low-power operation, since electro-optical conversions, analog-digital conversions and digital computations will no longer be needed to construct the output of the reservoir. To make this possible, also a novel training algorithm called non-linearity inversion was developed, which can deal with the incomplete observability of the reservoir states. All of this paves the way for larger reservoir which can tackle more complex problems. One important result of the PHRESCO project that signifies a progress beyond the state of the art is the experimental demonstration of non-linear dispersion compensation in an optical fiber line using a 1550nm reservoir with 32 nodes at record speed of 32 Gbps.
On the material level, the integration of barium titanate with strong electro-optical properties and a confirmation of the Pockels effect in active structures is clearly beyond state-of-the-art. Previously, only passive waveguides or active waveguides with ambiguous switching behaviour have been demonstrated. Bringing a material with strong Pockels coefficients into the silicon photonic platform enhances the opportunities for designers to create novel photonic circuits that could rely on physical effects previously not available in silicon photonics. In particular, our demonstration of integrating this material on wafers processed in a standard CMOS production line shows the applicability and scalability of the new technology. Moreover, the successful co-integration of the ferroelectric material and III/V technology on PIC platform demonstrated by the PHRESCO consortium by the end of the project is also beyond state-of-the-art.
On the level of photonic reservoir computing, the first generation prototype of Phresco goes beyond the state-of-the-art, since it will be the first realisation of a passive photonic reservoir with integrated all-optical readout. This will allow for high-speed low-power operation, since electro-optical conversions, analog-digital conversions and digital computations will no longer be needed to construct the output of the reservoir. To make this possible, also a novel training algorithm called non-linearity inversion was developed, which can deal with the incomplete observability of the reservoir states. All of this paves the way for larger reservoir which can tackle more complex problems. One important result of the PHRESCO project that signifies a progress beyond the state of the art is the experimental demonstration of non-linear dispersion compensation in an optical fiber line using a 1550nm reservoir with 32 nodes at record speed of 32 Gbps.
On the material level, the integration of barium titanate with strong electro-optical properties and a confirmation of the Pockels effect in active structures is clearly beyond state-of-the-art. Previously, only passive waveguides or active waveguides with ambiguous switching behaviour have been demonstrated. Bringing a material with strong Pockels coefficients into the silicon photonic platform enhances the opportunities for designers to create novel photonic circuits that could rely on physical effects previously not available in silicon photonics. In particular, our demonstration of integrating this material on wafers processed in a standard CMOS production line shows the applicability and scalability of the new technology. Moreover, the successful co-integration of the ferroelectric material and III/V technology on PIC platform demonstrated by the PHRESCO consortium by the end of the project is also beyond state-of-the-art.