Standard Multilayer Network Research Articles

Analysis–adjustment–synthesis networks

This paper presents a novel neural network architecture, analysis–adjustment–synthesis network (AASN), and tests its efficiency and accuracy in modelling non-linear function and classification. The AASN is a composite of three sub-networks: analysis sub-network; adjustment sub-network; and synthesis sub-network. The analysis sub-network is a one-layered network that spreads the input values into a layer of ‘spread input neurons’. This synthesis sub-network is a one-layered network that spreads the output values back into a layer of ‘spread output neurons’. The adjustment sub-network, between the analysis sub-network and the synthesis sub-network, is a standard multi-layered network that operates as the learning mechanism. After training the adjustment sub-network in recalling phase, the synthesis sub-network receives the output values of spread output neurons and synthesizes them into output values with a weighted-average computation. The weights in the weighted-average computation are deduced from the method of Lagrange multipliers. The approach is tested using four function mapping problems and one classification problem. The results show that combining the analysis sub-network and the synthesis sub-network with a multi-layered network can significantly improve a network's efficiency and accuracy.

Application of multilayered neural networks to ocean acoustic tomography inversions

Ocean Acoustic Tomography (OAT) is a synoptic observation method to infer the internal structure of the Ocean. The underlying idea is to estimate the three-dimensional sound speed distribution by inverting the travel times of a sound signal propagated between fixed sources and receivers. Usual inversion methods are derived from a linearization around a reference sound speed profile. This paper presents a non linear inverse method. This method relies upon the ability of neural nets to learn a non linear relationship from examples. A set of realistic sound speed profiles is built and the arrival time patterns in given experimental conditions are computed from a ray-tracing model. The inverse mapping is learned from this set of examples by a standard multilayered network. When unlearned time patterns are presented to the net inputs, it estimates the corresponding sound speed environments. An application of this method to synthetic data is presented in this paper. Four ocean model parameters (sound velocity v...