Progressive filter network

FIGURE 39. Progressive filter network for a binary classification of an unknown pattern x C173D. 

A much more sophisticated classification method with some similarities to the progressive filter network has been proposed by Meisel et. al. C2053. A set of decisions for a given classification problem is combined in a decision tree. Each terminal node of the tree indicates the probabilities for class 1 and 2. 

The KNN-method (Chapter 3) C2283 and the progressive filter network (Chapter 5.5) C1703 have been proposed for feature selection. 

During the filtration process cake progressively grows upwards from the filter medium surface, liquid is expelled from the press and the piston moves downwards in the direction of the medium. When the particles become sufficiently networked (which can be from the very start of the test), or enough cake has formed for the piston to touch the top of the cake, further deliquor-ing is achieved by consolidation where liquid is expelled as the porosity of the cake decreases. Consolidation continues until an equilibrium state is reached such that the particle structure/cake is sufficiently strong to withstand any tendency to compress under the applied piston loading. 

---