I preorderd Edward Tufte's Beautiful Evidence long before it came out, and devoured it as soon as it came. As always, it is probably the most typographically beautiful book that will come out this year. It just feels good sitting in your hand. It also has a strong message to tell. It is about how you should effectively present or consume all types of evidence. Despite that rather boring sounding topic, I found it fascinating.
As someone who works in Machine Learning (ML), I loved his discussion of building up cladogram trees showing evolution. These are very much like decision trees in ML. My favorite sentence in a long time is the second one here:
Sorting among many plausible candidate models often leads to publishing a brittle, over-fitted model. Thus these leaves may tremble. (p. 75)
(As an aside, ET's mother Virgina Tufte has written a book about good sentences. I wonder if that one will make it.)
I think all computer people should be required to read Tufte's books when they come out, both for their content and for an occasional dose of elegance.
If you've read his previous books, you'll find he does recycle his examples a bit. The graphic of Napoleon's Russian campaign is there again. Otherwise, a great book.
Comments