Saturday, December 6, 2014

Was that a blue bird? Ignoring observations during classification

I have a little story that illustrates possible approaches to handling false positives in a classification system with limited alternative categories. 

Driving in Bedford a small bird with a rust-red underbelly flies past the car and I get a brief (.3 sec?) glimpse.

Analysis: It was too small to be a blue bird and there are no birds that small around here with rust-red breasts. 
  • If the bird had been a bit bigger it could have been a blue bird or, even bigger, it could have been a robin.
  • If it wasn't really red then it could have been a Junco or sparrow.
  • If it was an accidental European bird, it might be one of those
  • If I misjudged size as well as position of the red color, it  could have been a Towhee or Catbird.
No small birds like that? Maybe the bird was a bit larger than I thought. For some reason I am more willing to dismiss a magnitude observation than a color one. Anyway, the mind obviously tries to loosen the observed 'measurements'; doubting each, in turn, in order to fit to a broader set of categories.

No comments:

Post a Comment