Timing in neural networks

Theme 01

Rhythms and entrainment

A network that already oscillates does not simply follow an external drive. How easily transcranial alternating current stimulation locks a recurrent E/I spiking network depends on the network's own coherence: weakly synchronised networks entrain over a wider band of frequencies, so their Arnold tongue is wider.

A second line asks what sets the rhythm in the first place. Inhibitory synaptic timescales drive transitions between oscillatory regimes in recurrent E–I networks.

Chaos, Solitons & Fractals 2025Physical Review Research 2026

Theme 02

Delays and plasticity

Connections between brain areas arrive with a spread of conduction delays. Spike-timing-dependent plasticity acting on those delays prunes the redundant connections and keeps the ones whose timing lets activity propagate from layer to layer, shaping networks that transmit signals efficiently.

Journal of Neuroscience 2026

Theme 03

Synchrony on networks

Kuramoto phase oscillators on a ring, each coupled to its nearest neighbours with delays that grow with distance, split into local synchrony and phase clusters. In two-dimensional spiking networks the same questions become questions about travelling waves.

kuramotoRing on GitHubTalk on phase-dependent suppression

Theme 04

Noise as a resource

Noise can help a neuron detect a weak signal. I study stochastic resonance in Hodgkin-Huxley neurons and networks with delayed feedback, comparing six circuit motifs from a single neuron to a balanced E–I network.

Notes on simulating noisy systems

Theme 05

Machine learning on brain data

Spatial-temporal Transformers and Conv-LSTM networks for forecasting seizures from multichannel EEG, and network analysis of EEG recordings in studies of human intelligence.