Abstract
Reservoir Computing is a computational paradigm employing the information processing capabilities of dynamical systems. This thesis investigates the echo state property, which is a state of generalised synchronisation between a reservoir computer's input and internal variables, using the concept of consistency. The consistency spectrum and capacity are introduced to quantify the degree and shape of this relationship, and to explain the performance of reservoir computers in unstable and noisy regimes. This is demonstrated with the development of a novel swarm based reservoir computer. The role of memory in reservoir computing is also shown through the example of coupled chaotic systems.
Original language | English |
---|---|
Qualification | Doctor of Philosophy |
Awarding Institution |
|
Supervisors/Advisors |
|
Thesis sponsors | |
Award date | 13 Oct 2021 |
DOIs | |
Publication status | Unpublished - 2021 |