Decaph
DECAPH DENDRITE-BASED COMPUTATION APPLIED TO PHOTONICS SYSTEMS

  • P.I.: Apostolos Argyris, Ingo Fischer, Claudio Mirasso
  • Coordinator: Ingo Fischer
  • Partners: IFISC-UIB
  • Start date: June 1, 2020
  • End date: May 30, 2022

The DECAPH project addresses fundamental aspects of cognitive computing introducing a disruptive architecture that mimics in more detail the actual operations that are performed in the brain. By mimicking such neuron structures in hardware platforms and investigating dendritic-based computational functionalities we plan to develop advanced cognitive computing schemes with the potential to tackle some of the crucial societal problems in a fast, flexible and energy-efficient manner.
Such concepts when transferred to high-performance technological platforms for computing may result in disruptive methodologies that may even alter the ways we perform currently cognitive computing and artificial intelligence assisted tasks. Photonic information processing has been revived lately with various techniques that offer computational capabilities related to time-series prediction, pattern
classification, and decision making. In the context of DECAPH, we will mimic optical dendritic structures inspired by a neuron dendritic tree - for addressing computing tasks. Optical fiber configurations offer the flexibility and technological maturity to implement such structures in hardware. Multiple spatially separated optical signals can be linearly or nonlinearly coupled, with on-demand induced time delays, using the previously mentioned fiber types. Therefore, photonic systems based on fiber-optics represent a promising and flexible platform for developing and implementing dendritic topologies for computing purposes. Employing single-mode-fiber-based topologies, optical dendritic units can be implemented by realizing multiple individual paths in order to build a coincidence-detection system. By exploiting the previous expertise on optical signal processing with fiber-based topologies, the IFISC groups will address the concepts of dendritic-tree computing, numerically and experimentally, initially on a unit level and eventually on a network level. The project itself targets
multidisciplinary goals that include: the expansion of the simple computational numerical models that describe the basic functionalities of a neuron by taking into account the dendritic structure including internal plasticity and nonlinear processes; the extension of the computational paradigm of a dendritic-based neuron unit to a network of units for complex functionalities on information processing; the
implementation of an optical computing platform based on single-mode fibers that will mimic a dendritic structure that will include also plasticity rules on the synaptic behavior of each dendrite; the implementation of a fiber-based dendritic-tree computing prototype with the ability to demonstrate adaptive integration functions and the potential to implement ultrafast optical information processing.

Researchers

  • Apostolos Argyris

    Apostolos Argyris

  • Ingo Fischer

    Ingo Fischer

  • Claudio Mirasso

    Claudio Mirasso

  • Moritz Pflüger

    Moritz Pflüger

  • Cristian Estarellas

    Cristian Estarellas

  • Irene Estébanez

  • Jaime Sánchez

    Jaime Sánchez

Recent Publications

Optical dendrites for spatio-temporal computing with few-mode fibers

Ortín, Silvia; Soriano, Miguel C.; Fischer, Ingo; Mirasso, Claudio R.; Argyris, Apostolos
Optical Materials Express 12 (5), 1907-1919 (2022)

Photonic neuromorphic technologies in optical communications

Argyris, Apostolos
Nanophotonics 11, 5, 897-916 (2022)

Microring resonators with external optical feedback for time delay reservoir computing

Donati, Giovanni; Mirasso, Claudio R.; Mancinelli, Mattia; Pavesi, Lorenzo; Argyris, Apostolos
Optics Express 30, 1, 522-537 (2022)

Information Transmission in Delay-Coupled Neuronal Circuits in the Presence of a Relay Population

Sánchez-Claros, Jaime ; Pariz, Aref; Valizadeh, Alireza; Canals, Santiago; Mirasso, Claudio
Frontiers in Systems Neuroscience 15, 705371: 1-16 (2021)

Predicting hidden structure in dynamical systems

Gauthier, Daniel J.; Fischer, Ingo
Nature Machine Intelligence 3, 281–282 (2021)

This web uses cookies for data collection with a statistical purpose. If you continue browsing, it means acceptance of the installation of the same.


More info I agree