To Do
- See Wunderlist
- Small changes to the data entry page
- Timestamps on the database for merging BYU data
- We could do a diff of their last db dump and the current one to get what changed
- Check on tuition payment next week
- Starting Monday
- Continue getting organized
- Reread papers, go over Bruno’s course notes, email Bruno
- Continue thinking about:
- equivalence classes,
- sliding window of identity definition (what gets considered to be an equivalence class),
- sliding window of time,
- continuous-time evolving networks.
Agenda
- Questions on funding (see Wunderlist)
- Any new news about the updated database from Shayne? (Getting MSSQL into Postgres)?
- Two weeks I’ll be out over the summer, if possible
- Likely June 13-21
- June 28-July 4 (ASP)
- Would like to work on proposing by end of summer
- Get document ready by June 28 (latest) and propose mid-July (latest)
- Still waiting on Wes for Math requirement
- Got a bill from student accounts, is that taken care of or should I deregister? (Due May 18)
- Go over the notes from the April 17 meeting with all, since they happened after Worthy left
Notes
- Interesting book to look up: Image Processing (textbook) by Rosenfield.
- Reel-time processing of videos (only have one shot as the reel goes across the head)
- Sounds like these current works are doing joint probabilities and/or using adjacency matrices
- We could have a big adjacency matrix, where each node is represented (even if it doesn’t exist at this particular time point). Then, as time progresses, nodes that show up will get edges and a weight.
- What’s missing: Nodes that change identity or merge
- Nodes (something special about them that an adjacency matrix won’t capture)
- Identity-defining characteristics don’t change OR they change the identity of the node
- Ie: a merger at a point in time (think back to the circus diagram)
- Other characteristics do change, though, which need to be captured about the node
- What about merging nodes?
- Can’t do adjacency matrix of all nodes that ever existed (as some groups do) because some are connected in a different way
- ie, a node at one time might be the same identity as two nodes at a previous time who merged
- Those nodes have a connection (linked identity across time), they are fundamentally connected, although different
- So, we have an ancestor-like connection
- It’s not an edge in the graph connecting two+ nodes at a given time, but it’s a connection at an identity level across time.
- Equivalence Classes of Nodes: we can pick a window of identity to determine equivalence classes of nodes over time.
- For the circus example, picking identity definition from 1919 will result in only one “type” of node all the way back, since the others merged in. This would show how the relations with “Ringling Bros and Barnum and Bailey” circus’ connections changed over time. However, picking an identity definition from 1881 would show 3 classes over time: Barnum and Bailey (the nodes that merge into this class in the past would be grouped together for all time), Ringling Bros, and then in 1919 something different would happen.
- The Mormon marriages are another story, which we’ve already discussed but using the term “dual” inappropriately.
- Let’s consider we have the marriage lineage flow, but where each node is only the binary marriages. So, this is the binary marriage diagram, in which there is only one husband and wife per marriage.
- We can set the identity window to group marriages identified by the man into an equivalence class, so we’d have an equivalence class per man in our database. This defines the patriarchal marriage diagram. All the marriages for a man are grouped together, and we then consider them one node in our network.
- However, we can also group by woman to get female-oriented equivalence classes and the matriarchal marriage diagram.