prev next

Probability Distributions

A certain number of example probability distributions are provided. Some are non-stationary, i.e. they change over time (time is not measured in seconds but in number of input signals presented).

Stationary Distributions

Most of the distributions are stationary i.e. they do not change over time. Examples are "Circle", "Corners" or "Cactus".

Non-Stationary Distributions

The non-stationary distributions are indicated by a name ending with "-N" like "Circle-N", "Corners-N" or "PulsingCircle-N". Only models with constant parameters, e,g, Growing Neural Gas, Growing Grid, or Hard Competitive Learning are able to handle non-stationary parameters. Models with decaying parameters, e.g. Self-Organizing Map or Neural Gas on purpose loose there adaptivity after a finite number of steps. Therefore the "Next PD" or "Prev PD" Buttons select only stationary distributions for these models.
One distribution - "DragMe" - allows the non-stationarity to be user-controlled using the mouse.

Rotation

Each stationary distribution (unless it is rotation-symmetric) can be turned into a non-stationary one by activating rotation