Occam's Razor

• recently, I've been embracing the idea that a simple black-box isn't necessarily better than an equally performant transparent utility that exposes hooks into the tools intermediates and allows the user to explore/exploit other aspects of the tooling.
• to summarize, I'd say the notions of complexity and simplicity can't be captured on a uni-dimensional scale where emphasizing one means sacrificing the other but they're instead orthogonal explorations that allow for independent tuning across different perspectives…

Given two equivalently performant solutions, employ the simpler one.

• Arises in the context of model complexity: Machine Learning and Deep Learning
  • a deep approach might not always be the right choice if one can do away with statistical methods/algorithms