Home > Summary of the project > Work packages > WP5 : Novel computing paradigms

WP5 : Novel computing paradigms

Friday 22 February 2013, by Alexandra Peereboom

Here, we target efficient (high speed or low power) analogue photonic or optoelectronic-computing systems based on the concept of reservoir computing (RC). This bio-inspired computing paradigm offers a framework to exploit the dynamics of analogue nonlinear dynamical systems directly for performing useful computation. It has been demonstrated to have state-of-the-art performance for a range of tasks that are notoriously hard to solve by algorithmic approaches, e.g., speech and pattern recognition and nonlinear control. In the previous phase of this IAP project the idea of using photonic systems as reservoirs was launched. Main results achieved so far are:
(i) Planar integrated photonic reservoir implementations were shown to match the performance of traditional (simulated) neural network based reservoirs.
(ii) A novel way of using the temporal dimension for performing analogue computation was proposed. This concept allows the use of low-dimensional systems with delayed feedback to operate as a high-dimensional reservoir.
(iii) An optoelectronic reservoir computer was built based on this concept. For several tasks, performance comparable to traditional reservoir computing was shown.

The new goals in the present project include:

5.1 Pushing back the digital boundary, where we will aim to perform as much of the readout and/or training in the optical, analogue domain, as opposed to the digital, electronic domain.

5.2 Exploring the limits of scale, speed and power consumption, where we will propose novel architectures to achieve ever better performance.

5.3 Targeting new and more complex tasks, where we will apply these photonic reservoirs to novel tasks, and draw up a roadmap for designing architectures suited to particular tasks.