Fast learning with incremental RBF networks

Abstract

Abstract. We present a new algorithm for the construction of radial basis function (RBF) networks. The method uses accumulated error information to determine where to insert new units. The diameter of the localized units is chosen based on the mutual distances of the units. To have the distance information always available, it is held up-to-date by a Hebbian learning rule adapted from the "Neural Gas" algorithm. The new method has several advantages over existing methods and is able to generate small, well-generalizing networks with comparably few sweeps through the training data.

Using CBR as Design Methodology for Developing Adaptable Decision Support Systems

“…Related with this domain we describe the FSfRT (Forecasting System for Red Tides) system, a hybrid model able to accurately forecast the concentrations of pseudo-nitzschia spp, the diatom that produces the most harmful red tides causing amnesic shellfish poisoning (or ASP). Our FSfRT system employs a case-based reasoning model to wrap a growing cell structures network, a radial basis function network (Fritzke, 1994) and a set of Sugeno fuzzy models (Jang et al 1997) to provide an accurate prediction. Each of these techniques is used at a different stage of the reasoning cycle of the decision support system to retrieve historical data, to adapt it to the present problem and to automatically review the proposed solution.…”

A biologically motivated visual memory architecture for online learning of objects

“…Incremental radial basis function (RBF) networks (Fritzke 1994a) and the growing neuronal gas (GNG) model (Fritzke 1995) were suggested with a focus on incremental learning. Although it is possible to train these networks with a slowly changing training set, these architectures are mainly designed for offline training.…”

A Hybrid CBR Model for Forecasting in Complex Domains

“…The GCS facilitates the indexation of cases and the selection of those that are most similar to the problem descriptor. The reuse and adaptation of cases is carried out with a Radial Basis Function (RBF) ANN [5], which generates an initial solution creating a forecasting model with the retrieved cases. The revision is carried out using a group of pondered fuzzy systems that identify potential incorrect solutions.…”

“…Initially, three vectors are randomly chosen from the training data set and used as centers in the middle layer of the RBF network. All the centers are associated with a Gaussian function, the width of which, for all the functions, is set to the value of the distance to the nearest center multiplied by 0.5 (see [5] for more information about RBF network).…”