Version Space Learning with DNA Molecules

Version space is used in inductive concept learning to represent the hypothesis space where the goal concept is expressed as a conjunction of attribute values. The size of the version space increases exponentially with the number of attributes. We present an efficient method for representing the version space with DNA molecules and demonstrate its effectiveness by experimental results. Primitive operations to maintain a version space are derived and their DNA implementations are described. We also propose a novel method for robust decision-making that exploits the huge number of DNA molecules representing the version space.

[1]  Thomas G. Dietterich.  What is machine learning? , 2020, Archives of Disease in Childhood.

[2]  Leonard Pitt,et al.  Version Spaces without Boundary Sets , 1997, AAAI/IAAI.

[3]  Masami Hagiya,et al.  Towards parallel evaluation and learning of Boolean μ-formulas with molecules , 1997, DNA Based Computers.

[4]  In-Hee Lee,et al.  Evolutionary sequence generation for reliable DNA computing , 2002, Proceedings of the 2002 Congress on Evolutionary Computation. CEC'02 (Cat. No.02TH8600).

[5]  Haym Hirsh,et al.  Generalizing Version Spaces , 1994, Machine Learning.

[6]  Yasubumi Sakakibara.  Solving Computational Learning Problems of Boolean Formulae on DNA Computers , 2000, DNA Computing.

[7]  John S. McCaskill,et al.  DNA Computing in Microreactors , 2001, DNA.

[8]  L. Adleman.  Computing with DNA , 1998 .

[9]  Nils J. Nilsson,et al.  Artificial Intelligence , 1974, IFIP Congress.

[10]  David Haussler,et al.  Quantifying Inductive Bias: AI Learning Algorithms and Valiant's Learning Framework , 1988, Artif. Intell..

[11]  David I. Lewin,et al.  DNA computing , 2002, Comput. Sci. Eng..