News for hackers. Stuff that matters.
Re:Back of the Envelope Calculations [pdf] by gus_massa on Tuesday April 09, 2013 @11:43AM (#5518887)
> A well-known curve of a calculation’s accuracy versus mental effort goes like

  \
   \
    \    __
     \__/  \
            \____ 
Have anyone really measured this bump or it's only folklore / myth?
Re:Back of the Envelope Calculations [pdf] by tinco on Tuesday April 09, 2013 @11:48AM (#5518921)
It's the well known 'this graph is true because it has unexplainable data in it' bump.
Re:Back of the Envelope Calculations [pdf] by gus_massa on Tuesday April 09, 2013 @12:47PM (#5519333)
The problem is not that it is unexplainable, or yet unexplained. The problem is if it is measurable and if it really exists.

(For example gravity is "unexplained" in some way. You have two masses, they deform the space, and with some approximations you can use the formula F=GmM/R^2. But, for example, why do masses deform the space?!?!?!?!?! But luckily you can measure it, just drop a stone or find a gravity lens, so gravity exists.)

Re:Back of the Envelope Calculations [pdf] by sesqu on Tuesday April 09, 2013 @11:26PM (#5523041)
It's certainly true in statistics and machine learning that adding variables with unreliable estimates lowers the whole model's accuracy after some point.