Research Article

Probabilistic Latent Variable Models as Nonnegative Factorizations

Figure 3

An example of a higher dimensional positive data decomposition. An isosurface of the original input is shown at the top left, the approximation by the model in (10) is shown in the top right, and the extracted marginals (or factors) are shown in the lower plots.
947438.fig.003a
(a)
947438.fig.003b
(b)
947438.fig.003c
(c)
947438.fig.003d
(d)
947438.fig.003e
(e)