Overview of Applications and Developments in the Harmony Search Algorithm - Music-Inspired Harmony Search Algorithm: Theory and Applications - page 25

where HMS is the Harmony Memory Size. As shown in Figure 4, the steps in the HS
algorithm are (i) initialising harmony memory; (ii) improvising a new harmony, that
is, generating a new candidate solution vector; (iii) updating HM with the new har-
mony if appropriate; and (iv) returning to step (ii) until some termination criteria are
satisfied. The improvisation of new harmonies in step (ii) involves three operations:
random playing, harmony memory considering, and pitch adjusting, as described in
Chapter 1. Several fixed-value parameters control these operations: the Harmony
Memory Considering Rate (HMCR); Pitch Adjusting Rate (PAR); and either the
bandwidth ( b ) for continuous variables or the neighbouring index ( m ) for discrete
variables, which are both used in pitch adjusting. If the new harmony is better than the
worst one currently in HM then HM is updated in step (iii) by replacing the poorest
solution with the new harmony. The algorithm terminates when it reaches a certain
maximum number of improvisations ( MaxImp ).

Begin HS
optimisation

Initialise HM

Improvise a new
harmony

Add new
harmony
to HM?

Y

Update HM

N

Termin-
ation criteria
satisfied?

N

Y

HS optimisation
completed

Fig. 4. Flowchart for the Original HS Algorithm