To Do
- Finish the abstract.
- Investigate the connected component claim of the 2003 paper that finding connected components in an evolving graph is NP Complete
- Is it NPC? What part of the temporal aspect makes it NPC?
- Would it still be NPC in our different temporal domains? Clearly there would be variation from a single point in time to an interval to multiple points in time.
- What is the definition of a connected component in a temporal graph?
- How would that change if we use the different notions of temporal domain (now, interval, disjoint interval, all time)?
- Would it have different meanings at different time points?
- Revisit the 2012 Temporal clustering paper
- What do they mean by cluster? temporal cluster?
- Look at the wikipedia datasets. Does an article node change over time? If the article is merged with another, is that reflected? Or if an article is split?
- Are there document-document edges along with editor-document edges?
Notes
- We have these different notions of time that we want to investigate. Different “places” to do our temporal measures:
- Now (single point in time)
- Contiguous Interval (interval with start and end points, inclusive)
- Non-contiguous Interval (set of intervals, with some points in time not included, possibly disjoint)
- All time (every point in time)
- For disjoint intervals, how do we look at the gaps and include edges or nodes?
- Do we include edges that are extant completely across the gap(s)?
- Edges or nodes that exist in all intervals included?
- Edges or nodes that exist throughout the gaps are included in the interval? That is, for an edge to be considered in a disjoint interval, it must continue to exist from the end of one interval through the beginning of the next to be considered.
- We need to look back at how vertices change over time
- We originally had a notion that connectedness is important based on some inherent characteristics of the node
- So, for example, we only want to consider circuses that have more than 3 travelling groups
- When a circus gains a third or loses its third group, the node changes importance to us. It changes whether or not we consider it, or how we rank it.
- Or, another example, we only want to consider wikipedia articles that have $x$ or more words from a specific list in them.
- When an article is edited and the $x$th word is added by the editor, that document is now more important to us, or important enough to include as a node in our graph
- When the $x$th word is removed so the article only has $x-1$ special words, we then treat that document differently
- Filtering vertices out that don’t have the properties we’re looking for.
- Alternatively treating them differently