We view perceptual tasks such as vision ad speech recognition as in- ference problems where the goal is to estimate the posterior distribution over latent variables (e.g., depth in stereo vision) given the sensory input. The recent flurry of research in independent component analysis exem- Plifies the importance of inferring the continuous-valued latent variables Of input data. The latent variables found by this method are linearly re- Lated to the input, but perception requires nonlinear inferences such as Classification and depth estimation.
展开▼