A coworker of mine has asked me for a term (preferably an adjective) that can be used to describe a system that gets more "intelligent" as it gets more data. The example she used when asking me this question was "as Google crawls more sites, it gets smarter at searching".
Personally, the best I could think of offhand was "adaptive", but that doesn't feel right. Can anyone suggest something better?
Thanks!
Sometimes you refer to things like spam filters as "trainable". Perhaps that could apply here.
The domain of this kind of applications is "machine-learning". But I'm not aware of a matching adjective.
It could be a vague description of an expert system, which often have a learning aspect and use it to gain more "expertise" in their problem domain.
The example she used when asking me this question was "as Google crawls more sites, it gets smarter at searching".
Unlike learning algorithms, where the algorithm itself changes based on past success, Google searches get better due to improved ranking of the results bringing the best pages to the top. The quality of the PageRank algorithm's results increases due to the network effect of the input data - the more connections, the better the chance that the best connected page is the most relevant.
The rule that says the effect of a network is super-linear is Metcalfe's Law, so if the "smartness" of an algorithm relies on network effects you could call the algorithm "Metcalfian". I've no idea whether the quality of PageRank results is super-linear in the number of inputs though; if anything I'd expect it to be sub-linear, as once you have enough links in the network to get rid of noise the rankings should be stable.
If it's a communications network, then it follows Metcalfe's law. You could call it Metcalfian. (You'd like be laughed at.)
I think the term is adaptive associative memory systems (leading to autonomy, perhaps).