Thursday 23 September 1999, S15, 3pm-5pm
Sequence Similarity.
Ad-hoc alignment methods:
Levenshtein metric, Longest Common Subsequence (LCS, LCSS),
2-D dynamic programming algorithm,
minimize some "weights" or maximize some "scores".
Most probable alignment, 1-state, 3-state etc. models of
sequence mutation / relatedness.
Sum of probabilities over all alignments;
average can be better (i.e. more appropriate)
than optimal!
See
TR-90/148
(Note relation to sum over all possible class-membership assignments in
Markov-Snob, as in last week).
NB. 27 Sept' - 1 Oct' is semester break.