“Improvement” Doesn’t = Up

Thu, Mar 18, 2010

Business, Philosophy

“Improvement” Doesn’t = Up

I’ve always taken “improvement” to mean up, rise, ascend.  On the ladder scale of “improvement”, anything upward moving is considered “good” and anything downward moving is considered “bad”.  For the sake of this argument, anything stationary is also considered “bad” because it’s not going up, which is the expected movement.

If we think about this though, “up” doesn’t always equate with “good” or “progress” or “better”.  “Up” could mean redundant, inefficient, or unproductive operations.  In fact, positive improvement could be down, lateral, or even on another axis.

There is not one direction or one way of achieving “improvement”.  As long as things evolve, in my eyes that’s improvement enough.

Just my two cents, but perhaps I’m crazy.

Photo credit swanksalot

, , , ,
  • In gradient descent search algorithms, it's often very important to have a way to perturb the current state out of a local maximum. If you define fitness (or improvement) only in terms of moving up, the computer will happily get stuck, king of its little hill, failing to notice the high value mountain just out of range.

    It's important therefore to not only have a method for deciding when you need to back down to go even further up, but to have a method for visualizing the higher possibilities outside your immediate sphere.

    (Yes, I will totally go computer science on this metaphor.)
blog comments powered by Disqus