Fig. 1. (a) Form of a neurule and (b) a neurule as an adaline unit.

Fig. 2. Neurule base production.

Fig. 3. The splitting tree for the merger set of Table 3.

Fig. 4. Architecture of the maintenance system of a neurule-based system.

Fig. 5. Insertion of R7 (I) (a) traversal of the splitting tree and (b) checking the leaf node at which traversing stopped.

Fig. 6. Insertion of R7 (II) (a) checking parent node’s merger and (b) final form of splitting tree.

Fig. 7. Removal of R5 (I) (a) traversal of the splitting tree and (b) checking the leaf node at which traversing stopped.

Fig. 8. Removal of R5 (II) (a) checking parent node’s merger and (b) final form of splitting tree.

Fig. 9. Removal of R6 (I) (a) traversal of the splitting tree and (b) checking the node at which traversing stopped.

Fig. 10. Removal of R6 (II) (a) checking parent node’s merger and (b) final form of splitting tree.

Fig. 11. Number of neurules for the cases of Table 10.

Fig. 12. Number of trainings for the cases of Table 10.

Fig. 13. Number of neurules for the cases of Table 11.

Fig. 14. Number of trainings for the insertions of Table 11.

Fig. 15. Number of neurules for the cases of Table 14.

Fig. 16. Number of trainings for the cases of Table 14.

A symbolic rule (R1) and a neurule (N1)

An example merger set of symbolic rules

The merger of the merger set of Table 2

Neurules produced from the merger set of Table 2

An example merger set of symbolic rules

Neurules produced from the merger set of Table 5

Inserted rule R7 and resulted neurule NR2-3-7

Merger sets and corresponding conclusions

General characteristics of the merger sets

Experimental results for rule insertions (SC1 and SC3 scenarios)

Experimental results for rule insertions (SC1, SC2 and SC3 scenarios)

Runtime results for the cases of Table 10

Runtime results for the cases of Table 11

Experimental results for rule removals

Runtime results for the cases of Table 14