The majority of literature on machine learning for resting-state functional magnetic resonance imaging (RS-fMRI) is devoted to unsupervised learning approaches. Modelling resting-state activity is challenging due to the absence of controlled stimuli driving fluctuations. Early analytic approaches focused on decomposition or clustering techniques to better characterize data in spatial and temporal domains. Unsupervised learning methods like ICA catalysed the discovery of resting-state networks or RSNs, which describe functionally coherent spatial compartments within the brain. Recent studies have shown that RSFC exhibits meaningful variations during a typical scan, making network dynamics even more interesting. The dynamic nature of functional connectivity opens new avenues for understanding the flexibility of different connections within the brain and their relation to behavioural dynamics.
