Experimental work across a variety of species has demonstrated that spontaneously generated behaviors are robustly correlated to variation in neural activity within the cerebral cortex. Indeed, functional magnetic resonance imaging (fMRI) data suggest that functional connectivity in cortical networks varies across distinct behavioral states, providing for the dynamic reorganization of patterned activity. However, these studies generally lack the temporal resolution to establish links between cortical signals and the continuously varying fluctuations in spontaneous behavior typically observed in awake animals. Here, we take advantage of recent developments in wide-field mesoscopic calcium imaging to monitor neural activity across the neocortex of awake mice. Diverging from traditional analysis of functional connectivity as a static entity, we explored the temporal dynamics of connectivity as expressed by instantaneous correlations between functional brain parcels. We develop a novel analysis, termed “graph-of-graphs”, that views the temporal fluctuations of correlations as high dimensional observations of a dynamical system and aims to extract their latent dynamics. We use Riemannian geometry and diffusion geometry to extract a low dimensional representation capturing the intrinsic dynamics of the functional connectivity. Using this novel approach, we demonstrate that spontaneous behaviors are more accurately represented by fast changes in the connectivity structure versus the activity of large-scale network. Moreover, the dynamics of the extracted functional connectivity representation reveals subnetworks that are not evident in traditional anatomical atlas-based parcellation of the cortex. For a small-scale network such as cells in the primary visual cortex, we show that there is no significant difference in the representation of behavioral variables using either embedded activity or embedded correlations, which means that the internal mechanisms of behavior encoding vary with scale. These results provide insight into how behavioral information is represented across the mammalian neocortex and demonstrate a new analytical framework for investigating time-varying functional connectivity in neural networks.