Work Package 1 - New concepts, learning & hardware optimization

Identification of critical gaps in the existing methods to perform learning in photonic systems.
Mathematical modelling of suitable methods for learning and optimization in analog photonic systems (development of novel concepts and/or combination of existing methods)
Testing of the developed methods in the hardware implementations developed by the POSTDIGITAL+ consortium
Create library of energy efficiency, complexity and convergence performance of various black-box optimization algorithms, continuous update through project.

Work Package 2 - Implementation and characterisation

Development of new concepts and techniques for non-algorithmic analog processing using photonics, including dynamical optical systems, solitons, and reservoir computing.
Full system implementation, including learning mechanisms, high-speed (GHz) implementations, and benchmarking against classical digital computing approaches.
Identification of possibilities for photonic integration to enhance system performance and scalability.
Training of DCs on analog computational concepts, system design, hardware implementations, and characterization of devices and systems.
Assessment of key influences, parameters, and noise in non-digital photonic processing concepts, ensuring robustness through alternative implementation concepts.

Work Package 3 - Integrated systems

Investigation of novel neuromorphic implementations based on integrated chips for high-speed, low-power computing.
Exploration of the standard silicon platform as a key candidate for integrating optical and electronic functions.
Expansion of previous work by UGent, ULB, HPE, Thales, and IBM to advance photonic neuromorphic computing.
Development of implementations suitable for industrial applications, including telecom, datacom, and image recognition.

Work Package 4 - Benchmarking and applications

Benchmarking of novel neuromorphic implementations on standardized machine learning tasks, including speech recognition, image recognition, and time series prediction, with respect to speed, energy consumption, and footprint.
Development of novel applications for neuromorphic implementations, including microwave equalization of nonlinear telecom communication channels (DC13).
Exploration of integrated optic accelerators for enhanced computational performance (DC14, DC15).
Investigation of novel ultra-fast real-time imaging technologies for advanced sensing applications (DC5).