Efficient Parallelisation of Learning Algorithms with Applications to Datamining

C. Rodríguez Lucatero (Mexico)


Computational Complexity, PAC-learning, parallel computing, data mining


In this article we show how can be parallelized efficiently PAC-learning algorithms for some specifically represented concepts that cover a very wide class of concepts to be learned in spite of the difficulty of parallelization of KDD based algorithms normally used in Data mining. Additionally we propose an alternative approach for doing KDD that makes a trade-off between performance and precision using parallel versions of PAC-learning algorithms for learning PAC-learnable concepts (concepts expressed in k-CNF[2][6] and monotonic k-DNF [14], simple decision lists [12][6], equivalence query simulation using less examples [13], logic recursive programs [15], concepts with finite Vapnik-Chervonenkis dimension [6])

Important Links:

Go Back