Samenvatting
Many of the natural phenomena we encounter in our daily lives are somehowinfluenced by delays. Traffic jams show regions of increasing and decreasing
speed due to delayed driver responses. Neuronal systems are influenced by the
delay associated with the distance between the neurons and the speed at which action potentials travels down the axon. Delay phenomena are not only interesting from the viewpoint of fundamental research. Increasingly nonlinear delay dynamics finds its way towards applications.
This work mainly focuses on experimental aspects and applications of delay dy-
namical systems. Specifically, we introduce a versatile and cost effective field programmable gate array (FPGA) platform, targeted towards tabletop experiments with delay differential systems. We analyze the dynamics of dual delay-coupled FitzHugh-Nagumo neurons with a Heaviside type nonlinearity. Also, we present a novel chaos-based encryption key distribution scheme implemented electronically, consisting of a ring of delay coupled nonlinear nodes, driving two matched slave systems into synchronization. The bitstreams of the slaves, which pass the NIST randomness test, are highly correlated, yet they share no correlation to the ring signals.
We investigate retriggerable and non-retriggerable monostable multivibrators as novel neurons in artificial neural networks. The age-old question of how to build a neuron in hardware is reversed, and we ask how a well known building block can be applied as a neuron. We show that these circuits, which are little more than a counter in digital form, have interesting dynamical properties when excited with stochastic pulse streams.
The concept of delay line based reservoir computing, using only a single non-
linear node with delayed feedback, was introduced some years ago as a means
of limiting hardware complexity in photonic systems. This concept turned out
to be beneficial for electronic implementations as well. On the FPGA platform,
we demonstrate a standalone delay line reservoir computer. It was given the
demonstrative task of predicting a chaotic system in real time. The influence of
noise and word bit depth on the performance was investigated using the same
setup. Finaly, replacing the delay differential equation at the core by a difference equation yields an efficient software structure, capable of real-time audio signal processing. This opens a much wider field of applications for delay line based reservoir computing.
| Datum prijs | 23 mei 2016 |
|---|---|
| Originele taal | English |
| Prijsuitreikende instantie |
|
| Begeleider | Guy Van Der Sande (Promotor) & Jan Danckaert (Promotor) |