Machine Learning, etc

Here's a side to side comparison of 150 vs. 300 vs. 600 DPI scan viewed at 400% magnification. On a local library scanner, 600 DPI black-and-white takes twice as slow to scan as 300 DPI, without significant improvement in quality

Newton once wrote to Hooke: "If I have seen further it is by standing on the shoulders of giants". It's true nowdays more than ever, and since there's such a huge volume of literature that is electronically searchable, the hard part isn't finding previous work, but remembering where you have found it.

Here's the strategy I use, which relies mainly on CiteULike and Google Desktop, what are some others?

• Use Slogger and "Save As" to save every webpage and pdf I look at, put the pdf's online for ease of access and sharing.
  
• For more important papers, add an entry to CiteULike with a small comment
  
• When common themes emerge (like resistance networks or self-avoiding walk trees), go over papers in that area and make sure they share a tag or group of tags
  
• For papers that are revisited, use "Notes" section for that paper in CiteULike to save page numbers of every important formula or statement in the paper.
  
• Finally, once a particular theme comes up often enough, review all the papers in that topic, write a mini-summary, put it in the "Notes" section of the oldest paper in that category
  

I do similar thing with books, in addition to scanning every technical book that I spend more than a couple of hours reading. With double-sided scans, you can average 4 seconds per page, so well worth the time investment. In addition, book scanning has a kind of meditation effect.

To find some result, ideally I remember the author or tag you put it under, then I use CiteULike search feature. If that fails, use Google Desktop to search through pdf's and web history.

What strategy do you use?

Pi comes up in the most unexpected places. Here's an application to walk counting that involves it.

Suppose you have a chain of length n. How many walks of length k are there on the chain? For instance, for a chain of length 5, there are 5 paths of length 0 (start at each vertex and don't go anywhere), 8 of length 1 (traverse each edge either left-right or right-left), 14 of length 2 (8 walks that change direction once, 6 walks that don't change direction) etc.

Curiously enough, there's an explicit formula for this



For example, to find the number of walks of length 2 in a chain of length 5, plug n=5, k=2 into formula above, and get



Which is 14.

Notebook, web version