Inspired by biology, neuromorphic (brain-like) computing aims to capture the incredible capabilities of the human brain with physical devices. The brain completes numerous complex tasks such as pattern recognition using a fraction of the time and energy of conventional computers. In order to build systems with similar computational abilities, it makes sense to investigate systems that have intrinsic brain-like qualities.

Percolating networks of nanoparticles (PNNs) have many features that make them suitable for neuromorphic computing. The neuromorphic properties of PNNs have previously been demonstrated for simple two-electrode devices. Multiple electrodes are required, however, for the implementation of PNNs as the reservoir in a conventional reservoir computing (RC) scheme, where multiple inputs and outputs are used. This thesis focuses on simulations of PNNs and it is demonstrated that the neuromorphic properties are conserved in devices with multiple electrodes.

RC utilises temporal correlations in a dynamical system in order to perform computation. A model for operating in the low-voltage tunnelling regime is described, which allows PNNs to be used as the physical reservoir. A range of benchmark tasks are successfully performed and the effect of network size on RC performance is investigated.

The RC scheme is extended to the emulation of swarming behaviour, like that seen in flocks of birds. A number of approaches are investigated and a novel method is developed that produces ‘swarm-like’ behaviour with PNNs for the first time.

The results presented in this thesis demonstrate that PNNs have capability for neuromorphic computing and show that there is promise for solving even complex problems.