Abstract data processing model (2)
Immediate examples are the algorithms that apply LDA after having performed PCA.
Many well-established algorithms encompass an expressive step (i.e., an approximation step) and a discriminating step (i.e., a detection step):
- Vector quantization is composed of an optimal encoder for a given decoder and an optimal decoder for a given encoder.
- Adaptive subspace self-organizing maps, radial basis function NN, the expectation-maximization algorithm follow the same paradigm.
An additional property is desirable, namely, simplicity.