interfacing a photonic reservoir computer with a high-speed electronic device, a
Field-Programmable Gate Array (FPGA), enabling interactions with the reservoir
computer in real time.
The main results obtained during the thesis are:
(1) The experimental demonstration that photonic reservoir computers can solve
tasks that change in time, which is important for some appli cations, particu-
larly in telecommunications. For this, Piotr had to adapt the method known as
online training to photonic reservoir computers.
(2) The experimental demonstration of a reservoir computing system whose
hardware is used both for computing and for optimising its internal parame-
ters, thereby considerably improving performance. By carrying out the opti-
misation in hardware, this work demonstrated that one could potentially
circumvent the difficulties of previous work that relied on detailed numerical
models of the experiment to realise the optimisation.
(3) The experimental demonstration that photonic reservoir computers with output
feedback could produce periodic time series and emulate chaotic dynamical
systems. This opens up a whole new area of applications of analogue reservoir
computers, as well as novel conceptual questions about chaotic dynam ics.
(4) The numerical demonstration (which Piotr Antonik hopes to soon confirm in
hardware) that the performance and robustness of the output layer of reservoir
computers could be considerably improved by online training.
(5) Real-time image analysis of optical coherence tomography of atherosclerotic
arteries, using methods from artificial intelligence and FPGA programming.
(In this chapter, Piotr Antonik used his expertise in machine learning and
FPGA programming to make progress on a very different problem, namely
real-time image analys is of optical coherence tomography of atherosclerotic
arteries).
The thesis may give rise to industrial valorisation. First, reservoir computing is
simple, fl exible, easy to train and can be implemented in optics. It is believed that
the most promising area of application in the short term is telecommunication
through optical fibres, and in particular equalisation of optical communication
channels (i.e. corrections of the distortions that occur during transmission). Second,
Piotrs work on optical coherence tomography may give rise to applications on a
short timescale. Indeed real-time image analysis of optical coherence tomography
of atherosclerotic arteries may help expand the use of this technique in hospitals.
Piotr Antoniks thesis is highly interdisciplinary. The theory on how to design,
train, exploit and benchmark the experiments was supplied by ideas from machine
learning, and in particular from the topic of reservoir computing. The experiments
themselves were built from photonic and electroni c components. Understanding
how they would perform required extensive numerical simulations. To develop new
features, Piotr used an advanced tool from the electronics world—a FPGA
chip. Programming of such a chip is a complex task that requires in-depth expertise
of analogue and digital electronics.
