An experimental study on the modeling ability of the IDS method

The ink drop spread (IDS) method has been proposed as a modeling technique for the active learning method. The IDS method, which uses pattern-based processing instead of complex formulas, is able to deal with various modeling targets, ranging from logic operations to complex nonlinear systems. Although the computing structure of the IDS model is characterized by heavy parallel processing on distributed units, its modeling process is simple and efficient and does not require iteration of the same training data set observed in the learning of neural networks. This paper experimentally studies the modeling ability of the IDS method through some typical benchmarks.