The effects of the message length statistics on the mean message response time and message and packet access delays are investigated for variable-length multipacket messages transmitted over CSMA-CD (carrier-sense multiple-access with collision detection) local area networks. The steady-state distribution of the number of messages in the system is determined using the imbedded Markov chain technique and limiting results from renewal theory. Application of Little's law then yields the mean message response time, from which the mean message and packet access delays are determined easily due to the gated transmission strategy adopted. Sample numerical results are presented for geometrically and truncated Poisson-distributed message lengths. It is concluded from the numerical results that the access delays as well as the mean response time are sensitive not only to the message length statistics but also to the message length distribution and the node buffer size.>

An algorithm for pattern classification using eigenvectors is presented. This algorithm provides an alternative viewpoint for obtaining a solution weight vector of certain pattern classification problems. Convergence of the algorithm is proved and the termination condition of iteration is given. An interesting point is that this algorithm does not require the inversion of matrices, instead it requires the computation of eigenvalues and eigenvectors of the matrix.