Research Stability of Parallel Algorithm for Solving Strong Separability Problem Based on Fejer Mappings

The problem of strong separating has an important role in the pattern recognition theory. The problem of strong separating means separating two convex non-intersected polyhedrons by the layer of maximum thickness. In this article, the non-stationary problems of strong separating are considered. Non-stationary problem is a problem for which the input data have been changed during the calculation process. An algorithm solving the non-stationary problem of strong separating must have two properties: auto-correcting and stability. Auto-correcting means the algorithm can continue its work effectively after the input data have been changed. Stability implies a small input data change implies a small deviation of the result. The auto-correcting is the feature of iterative algorithm based on Fejer processes. In the paper, the parallel algorithm based on Fejer mappings is described. This algorithm admits an effective implementation for the massively parallel multiprocessor systems. The notion of stable Fejer mapping is introduced. The theorem about stable Fejer mapping is proved.

Keywords

Fejer mapping, problem of strong separating, iterative method, pseudoprojection of point, stable Fejer mapping.

References

1. Eremin I.I. Feyerovskie metody sil'noy otdelimosti vypuklykh poliedral'nykh mnozhestv [Fejer Methods for the Strong Separability of Convex Polyhedral Sets]. Russian Mathematics, 2006, no. 12, pp. 33 - 43.
2. Boser B., Guyon I., Vapnik V. A Training Algorithm for Optimal Margin Classifiers. Proc. of the 5th Annual ACM Workshop on Computational Learning Theory. Pittsburgh, ACM Press, 1992, pp. 144 - 152.
3. Eremin I.I. Nestatsionarnye protsessy matematicheskogo programmirovaniya [Nonstationary Processes of Mathematical Programming]. Moscow, Nauka, 1979. 288 p.
4. Ershova A.V., Sokolinskaya I.M. Parallel'nyi algoritm resheniya zadachi sil'noy otdelimosti na osnove feyerovskikh otobrazheniy [Рarallel Algorithm for Solving Strong Separability Problem Based on Fejer Mappings]. Vychislitel'nye metody i programmirovanie [Numerical Methods and Programming], 2011, vol. 12, no. 2, pp. 53 - 56.
5. Eremin I.I. Teoriya lineynoy optimizatsii [Theory of linear optimization]. Ekaterinburg, 'Ekaterinburg', 1999. 312 p.
6. Ershova A.V. Algoritm razdeleniya dvukh vypuklykh neperesekayushchikhsya mnogogrannikov s ispolzovaniem feyerovskikh otobrazheniy [The Algorithm of Separation of Two Convex not Intersect Polyhedral Using Fejer Mappings]. Sistemy upravleniya i informatsionnye tekhnologii [Controlling System and Information Technology], 2009, no. 1(35), pp. 53 - 56.
7. Chernikov S.N. Lineynye neravenstva [Linear inequalities]. Moscow, Nauka, 1968. 488 p.