1) Temporal networks are networks where elements like ties and node properties change over time, unlike static graphs. They require new analysis methods that account for temporal dynamics.
2) Visualization techniques for temporal networks include static aggregated graphs, weighted aggregated graphs, and timelines. Studies of citation patterns between journals show how disciplines evolve over time.
3) Formulating temporal networks involves defining time windows and finding the graph state at random times within or across windows. Metrics are adapted from static graphs to measure temporal aspects like reachability and centrality over time.
2. Temporal Networks ο Networks in which Elements Change Over Time
β’ Fluctuation in Tie Weight
β’ [Mainly] Intermittent Tie
β’ Actors Property Values
Time-Evolving | Temporal | Time-Varying | Dynamic Graph β Static Graph
DEFINITION
3. VISUALIZATION
β’ Static Graph
β’ Aggregated Static Graph: At least one connection during the time span
β’ Weighted Aggregated Graph: Weigh the connection persistency by probability or other measurements
β’ Timeline
4. Studying the citation pattern between about 7000 scientific journals over the past decade
Neuroscience from an Interdisciplinary Specialty ο Mature & Stand-alone Discipline
Alluvial Diagrams
6. Time-Respecting Walks
Temporal Walk from Node i to Node j is a Time Increasing Ordered Sequence of L Edges
νν0 , νν1 , νν1 , νν2 , β¦ , νννβ1 , ννν
νν0 = ν, ννν = ν
ν‘ν1 < ν‘ν2 < β¦ < ν‘νν
ννννβ1
METRICS
Distance
,ννν
ν‘νν β 0
Topological: Number of Edges Traversed by the Path
Temporal: Time Interval or Duration between the First & the Last Nodes
Reachability, Connectedness, Centrality(Betweenness, Closeness, Spectral), β¦
7. TEMPORAL SCALE
Interval of Time ο A Minute, Day, or A Year
β’ kinship relations!
Oversampling
β’ Affect the Ability to Distinguish the Change
β’ Technology ο Very Fine Grained Snapshots
β’ Problem under Study ο Determine the Scale
β’ Tweeter: Minute
β’ Social Tie: Month
Aggregation
β’ Increase the Time Interval while Preserve Information
Heuristics: Persistence is a property that allows us to construct a network with the βcoreβ interactions, discarding the
noisy transient interactions. βrightβ temporal scale: the temporal scale that best captures the persistent nature.
β’ TWIN: Temporal Window In Networks
β’ Graphscope
8. The TWIN (Temporal Window In Networks) heuristic uses graph-theoretic measures as proxies of different aspects of network structure.
Given a temporal stream of edges and a graph-theoretic measure, the heuristic generates time series of graphs (dynamic graphs) at different
levels of aggregation. It then computes the variance and compression ratio for each time series. Finally, the algorithm analyzes the
compression ratio and variance as functions of window size and selects the window size for which the variance is relatively small and
compression ratio is relatively high.
Graphscope uses the notion of compression cost to capture the persistence of network structures (in this case of communities) in time.
Similar graph snapshots will incur low compression cost, therefore they can be grouped together in one temporal segment. Whenever
the compression cost increases substantially with the addition of a new graph snapshot, Graphscope starts a new temporal segment.
A nice feature of the Graphscope
heuristic is the fact that it generates a
nonuniform partitioning of the timeline.
The non-uniform partitioning is a more
realistic representation of real-world
interaction streams which are
commonly characterized by bursty
behavior
9. Predicting the Temporal Dynamics of Information Diffusion in Social Network
β’ Learning Target Function 0 β€ νν₯,ν¦ ν‘ β€ 1by 4-D Feature Space of {User, Topic, Topology, Time} ο {Diffused or Non-Diffused}
Spatio-Temporal Dynamics of Online Memes: A Study of Geo-Tagged Tweets
β’ Hashtag Adoption Lag
β’ Measuring Spatial Impact
APPLICATION
10. β’ Spatio-Temporal Dynamics of Online Memes: A Study of Geo-Tagged Tweets
β’ Peak Analysis of Hashtags & Relation between the Pace to Reach the Peak (fast or slow) to the Spatial
Distribution
β’ The Peak of Hashtags Propagation in terms of Occurrences