Welcome to the HPCA 2015 dining excursion

HPCA will be catching dinner (depending on the restaurant, starting at 5:45, 6:00, or 6:15) and then a show!

You should shoot to arrive at Beach Blanket Babylon at 8:15p. You do not need to come earlier than this, and you should not arrive much later either.

Prof. Michael Taylor, UC San Diego
General Chair HPCA 2015

We collected data from all registrants up to January 26, 2015; which included preferences of up to six people to sit with, whether price was more important than who you sit with, as well as institution and position.

Our algorithm

After extensive data scrubbing (e.g. Mike v. Michael, Joe v. Joseph, UC San Diego v. UCSD, etc), we created a directed graph (where an edge denotes "A wants to dine with B") and applied the following algorithm:

  1. First, we discarded all edges from graduate students to people at the same institution, including both advisors and students.

    We did this because graduate students always need a little bit
    of incentive to network with other people at the conference.
    In fact, after we grouped everybody, each participant was emailed a link with a list of the
    people they are grouped with so that they can look up all of the people
    in your dining group and see who they could have an interesting conversation with!

    Check out their websites and Google Scholar pages.

  2. Second, we took the small number who selected the "cheap" option, and put them in a group.

    They will go to a legendary hole-in-the-wall called Golden Boy Pizza that serves excellent pizza by the slice, and beer.
    But it's tiny and you have to sit at the bar, or stand in a group. It's one dollar sign (out of four) on yelp.
    If you are in a cheap group but indicated that you just wanted to be placed with your friends,
    your friends may have sucked you in. =)

    Note: we assumed that graduate students and postdocs have a
    different definition of cheap than US/Europe-based professors and industry;
    for the latter group, we made sure that all of our selections were two dollar
    signs (out of four) on Yelp and are relatively affordable
    for the area, so that we could mix these people freely without regard to cost.

  3. Third, to ensure we seeded the restaurants with a good mix
    of industry, gov't, and academic leaders, we partitioned the large
    connected subgraph of non-graduate students into three
    major components, corresponding to the three large restaurants we
    had tables reserved at.

    We made sure that we never cut the only outgoing arc of any given node;
    i.e., if a person listed at least one person who is coming to the dinner
    they will be grouped with at least one such person.

  4. We sprinkled singleton nodes across the graph to ensure a good mix of industry and academia at each location.

  5. Finally, saving the best for last, we added all of the graduate students
    and undergraduates according to the preference edges.
    Here is the HPCA social dining graph!