Estimation of ocean-sediment properties by sparse noisy measurements u
sing generalized Radial Basis Function networks is proposed. Given a s
et of scattered data points, an RBF network is able to generate a cont
inuous smooth approximation for sediment properties as a function of t
he x-y-z position, where z is the sediment depth. Advantages and disad
vantages of the method are discussed, from both a physical and a compu
tational viewpoint. An example using sediment density data obtained by
sparse core measurements and different configurations of RBF networks
is presented.