10 Matching Annotations
  1. Sep 2018
    1. numbers


    2. and its distance to all points on its left willincreasebyδ

      This relies on an unstated fact that the shortest path between these points is along this line. I don't know how familiar you're expecting the reader to be with this.

    3. the line

      "points on a line"

    4. But a high-quality solution is not the same thing as a meaningful solution

      Now we're talking

    5. tor of 2 is NP-hard, and is therefore as hard as solving the problemexactly

      At this point I wonder: "How does one then distinguish among the many possible solutions that are 2-approximations?"

    6. space of partitionsΠ(X), we can search oversubsets ofXof sizek

      Perhaps a note on how much smaller the latter is?

    7. Effectively, we are searching for a partition in which the maximum radiusof a cluster is minimized.

      It's been a few pages since jumping into definitions. It may be worthwhile to emphasize somewhere earlier that these particular definitions are not applicable to every clustering problem, but rather it's a set of examples that cover a large set of techniques while showing many of the important considerations.

    8. help guide how we think about clustering a data set

      Now I'm hooked

    9. “related”.

      Should these quotes be `` (double backtick) '' (double apostraphe)?

    10. collection