Matching time‐lapse seismic data using neural networks

In this paper we use artificial neural networks to crossequalise one vintage of time-lapse seismic data to another. The networks act as three-dimensional, non-linear operators, and are therefore capable of correcting for both spatial and temporal mismatches; indeed, the mismatches can also be non-stationary. In general, the networks are described by a relatively small number of parameters, and this in turn helps maintain stability.