I recently became aware of a bridge between (dynamical) properties of time-series and (topological) features of an associated network representation. A variety of methods exist to embed the time-series into a network (see e.g., Transforming Time Series into Complex Networks by Michael Small, Jie Zhang and Xiaoke Xu).
As far as I understand, to each univariate time-series is associated a single network. I wonder if you are aware of a method to embed a set of univariate time-series into a single network.
Yes, there are so-called functional networks, where you associate each time series with a node in the network. You then determine the existence or weights of edges by applying interaction measures to the respective pair of time series. The most simple such interaction measure would be the correlation coefficient, but you can also apply information theory, phase reconstructions, etc. There is an entire subfield dedicated to properly deducing interactions from time series.
This technique has been applied in several fields, most prominently brain and climate science, where networks have been deduced from time series measured with sensors placed on or in the brain or on earth respectively.
I am not aware of a recent good overview focussing exclusively on this, but this and this paper provide some starting point.