3 ms·
A similar strategy is to find things which were invented 10–100 years ago but considered insignificant then, but due to changing environments can be adapted to
by kqr 4y ago
A similar strategy is to find things which were invented 10–100 years ago but considered insignificant then, but due to changing environments can be adapted to be useful today in a way that wasn't possible then.
My day job is to work on software working with self-organising maps, which were last popular in 1980. As far as I know, they were never really used for anything significant then. The reason we can be so successful using them is that we discovered a relatively new environment in which they can improve profits fairly immediately: e-commerce.
- jasonb05 4y agoMan I love SOMs. When Kohonen's book came out in 2000, I read that thing a bunch of times. Seems like an expensive way to structure a "this is like" or "people also bought" compared to an SVD.
- kqr 4y agoThe reason we prefer SOMs is that we have been able to get them to produce good results with less data than e.g. an SVD. They're also more open to domain expert diagnostics and tweaking.