The problem under consideration is that of combining classification results from several classifiers. This problem was motivated by the US Navy's involvement in developing a system for integrating classification results from multiple sensors. As in any decision process one needs to quantify the uncertainty associated with the decision. A procedure based upon ranked lists from classification outputs is considered. Particular attention is given to the problem of assigning measures of confidence to the combined classification results where the interpretation is easily explained.>

The Frank-Heimiller sequence of period N2withNphases is constructed by writing theNbyNdiscrete Fourier transform matrix by rows. The periodic autocorrelation sidelobes for such a sequence are all zero. Small families ofN-phase sequences, with optimal periodic cross correlations, can be formed by judicious selection of decimations of the Frank-Heimiller sequence.

Five binary linear[n, k, d]codes are presented that contain more words than the codes with corresponding values ofnanddthat are listed in the table of MacWilliams and Sloane. The construction method is a generalization of ConstructionXof Sioane, Reddy, and Chen.

Three types of families of complex periodic sequences are shown to have nearly minimal correlation magnitudes. The sequences contain no zero entries. For certain pairs of sequences of periodN=(p- 1)/2, pa prime, all nonpeak correlation coefficients have magnitude close to(2N)^{-1/2}.

A recent table of Helgert and Stinaff gives bounds ford_{max}(n,k), the maximum minimum distance over all binary linear(n,k)error-correcting codes,1 \leq k \leq n \leq 127. Twelve new codes are constructed which improve lower bounds in the table. Two methods are employed: the algebraic puncturing technique of Solomon and Stiffler and generation by combinatorial incidence matrices.

An algorithm is presented for transposing large nonsquare matrices stored externally or in core. A fast algorithm of Eklundh and direct algorithm are applied to partitioned matrices and combined to effect the transpose of a nonsquare matrix. The matrix must be augmented so that its dimensions have a large common divisor. In the case of large externally stored matrices, additional external storage for a matrix of equal size is not required.