intheTuringmodel.PDFVIP

On the complexity of some computational problems in the Turing model Claus Diem November 18, 2013 Abstract Algorithms for concrete problems are usually described and ana- lyzed in some random access machine model. This is in particular the case in the areas such as computational algebra, algorithmic number and cryptology. In this work the complexity of various computational problems is studied in the multitape Turing model of computation. It is shown that “up to logarithmic factors” key results also hold in the multitape Turing model. Speci?c problems which are considered are linear algebra computations, factorization of integers and the compu- tation of discrete logarithms in various classes of groups. 1 Introduction The area of computational complexity falls roughly into two parts: On the one hand we have “abstract” complexity theory and on the other hand, we have the study of algorithms for particular computational tasks; cf. [San12]. In the ?rst part, exponents in running times are not considered and therefore it does not matter if an algorithm is analyzed on the basis of a particular RAM model with logarithmic cost function (and with a reasonable set of instructions) or in the Turing model of computation. On the other hand, if one studies algorithms for particular computational tasks, often the ex- ponents are of crucial interest and consequently it is important that it is clearly stated what the underlying machine model is. This work is concerned with the algorithmic side of computational com- plexity. Here, most of the time one chooses a particular random access machine (RAM) model as a formal basis for the analysis of an algorithm. In particular, it is generally postulated that almost immediate transfer of data from whatever location of the memory to a central processor is pos- sible. For exam