The Variational Correlation Network for Object Detection

We present a variational correlation network that can well represent marginal distribution as well as correlation among the sites. The network is represented by the complex valued equations, which consist of phase equations and variational mean-field equations; thus the correlation coefficient between two sites on stochastic machines can be represented by the cosine of the phase differences. This enables us to compute the correlation between two units directly in a deterministic manner. The variational correlation network improves the accuracy of the mean field approximation for generative models. Unlike the conventional elaborated mean field methods the efficient training method can be implemented on this model. The variational correlation network proposes a much more efficient learning model for pattern detection or image processing than Markov network models