4 ms·
My nickname for this phenomenon is the "robustness/efficiency spectrum." In this case, it would be the robust generalists and the efficient specialists on oppos
by sidpatil 6y ago
My nickname for this phenomenon is the "robustness/efficiency spectrum." In this case, it would be the robust generalists and the efficient specialists on opposite ends of the spectrum.
(I'm not sure it has to be a two-dimensional spectrum in all scenarios —I believe there may be cases in which robustness and efficiencies are at least partially orthogonal/independent to each other.)
Taleb discusses this in his writing, if I recall correctly.
- mratsim 6y agoIn reinforcement learning which is used for things like AlphaGo, there is a similar duality called "exploration vs exploitation": - do you keep searching in breadth for new avenues - or is it time to dive in depth in what you found so far. And this is even optimized through something called "regret". - example: https://rlss.inria.fr/files/2019/07/exploration_lazaric_1.pdf https://rlss.inria.fr/files/2019/07/exploration_lazaric_1.pd...