3 ms·
It makes me wonder - is there a good hill-climbing algorithm that takes account of uncertainty - where the hills can go up and down over time as you're climbing
by ojbyrne 12y ago
It makes me wonder - is there a good hill-climbing algorithm that takes account of uncertainty - where the hills can go up and down over time as you're climbing them? I suspect the naive algo might perform better in those conditions - maxima that are closer to you would be favored.
- placebo 12y agoSimulated annealing was already mentioned. Also parallel tempering is very effective at finding the highest (or lowest) areas even faster. I actually liked the analogy but applying algorithm analogies to one's life of course only goes so far. First, even if one decides to look at life as an optimisation problem, it is much more difficult than the hardest multiple objective optimisation problems. One of the reasons is that I'd equate the optimisation landscape not with static hills, but more similar to waves or sand dunes that keep shifting with time in a way that can at most give you short term predictions. We don't have the life span or processing power to use a computer algorithm to optimise a static "life" landscape, let alone a dynamic one.