Semi-Markov  chains  are  used  for  studying  the  evolution  of  seismicity  in  the Northern Aegean Sea (Greece). Their main difference from the Markov chains is that   they  allow the sojourn times (i.e. the time between successive earthquakes), to follow any arbitrary distribution. It is assumed that the time series of earthquakes that  occurred  in  Northern  Aegean  Sea  form  a  discrete  semi-Markov  chain. The probability law of the sojourn times, is considered to be the geometric distribution or   the   discrete  Weibull  distribution. Firstly,  the  data  are  classified  into  two categories that is, state 1: Magnitude 6.5  -7 and state 2 Magnitude>7, and secondly into three categories , that is    state 1: Magnitude 6.5-6.7, state 2: Magnitude 6.8-7.1 and state 3: Magnitude 7.2-7.4 . This methodology is followed in order to obtain more accurate results and find out whether there exists an impact of the different classification on the results. The parameters of the probability laws of the sojourn times are estimated and the semi-Markov kernels are  evaluated for all the above cases  .  The  semi-Markov  kernels  are  compared and  the   conclusions  are  drawn relatively to future seismic hazard in the area under study.

No abstract

In this paper we present a new Viterbi algorithm for Hidden semi-Markov models and also a second algorithm which is a generalization of the first. These algorithms can be used to decode an unobserved hidden semi-Markov process and it is the first time that the complexity is achieved to be the same as in the Viterbi for Hidden Markov models, i.e. a linear function of the number of observations and quadratic function of the number of hidden states. An example in DNA Analysis is also given.