Material Handling: A Conservative Domain for Neural Connectivity and Propagation

Abstract

Two important components of connectionist models are the connectivitV between units and the propagation rule for mapping outputs of units to inputs of units. The biological domains where these models are usuallV applied are nonconservative, in that a single output signal produced bV one unit can become the input to zero, one, or manV subsequent units. The connectivitV matrices and propagation rules common in these domains reflect this nonconservativism in both learning and performance. CASCADE is a connectionist sVstem for performing material handling in a discrete parts manufacturing environment. We have described elsewhere the architecture and implementation of CASCADE [PARU86a] and its formal correspondence [PARU86c], [PARU87a] with the PDP model [RUME86]. The signals that CASCADE passes between units correspond to discrete phVsical objects, and thus must obeV certain conservation laws not observed bV conventional neural architectures. This paper brieflV reviews the problem domain and the connectionist structure of CASCADE, describes CASCADE's scheme for maintaining connectivitV information and propagating signals, and reports some experiments with the sVstem.