3 ms·
The problem with the NFL theorem is that, in a sense, it doesn't apply to the real world: https://en.wikipedia.org/wiki/No_free_lunch_in_search_and_optimizatio
by cgmg 9y ago
The problem with the NFL theorem is that, in a sense, it doesn't apply to the real world:
https://en.wikipedia.org/wiki/No_free_lunch_in_search_and_optimization#NFL_and_Kolmogorov_randomness https://en.wikipedia.org/wiki/No_free_lunch_in_search_and_op...
- viewtransform 9y agoTrue. High-dimensional non-linear systems (like deep NNs) are a frontier we are just beginning to explore with empirical success. However, if I was a VC - my knowledge of the NFL would make me look skeptically at any grandiose claims of miracle learning algorithms. In a sense, it provides me with an upper bound to BS.