[bdtcoapd] gaussian mixtures

consider a game of wandering around a universe populated with multivariate gaussian distributions with different centers and parameters.  each point in the game is described by a large (perhaps infinite) number of features, each feature a multivariate gaussian of position.

(these features are time-independent scalar fields, in physics terminology.)

find a point where features overlap in a desirable ratio, perhaps input resources for doing something at that spot.  as you progress, you acquire better tools to detect fainter signals from more distant gaussians, to climb gradients and follow contours, to make maps.

perhaps substitute one resource for another, at a cost.