Real-World Applications - Immunological Computation: Theory and Applications - page 188

eff ector cells and combines both positive and negative detection; it also keeps
track of the concentration of two cytokines in the environment. It is based on the
assumption that there must be an interaction between cells in the population before
determining whether an antigen belongs to self or nonself. h e system possesses a
memory that is represented by cytokine concentrations such that the classifi cation
of an antigen depends on the responses against recently classifi ed instances. h e
system does not include clonal selection, thus the memory is not antigen specifi c.
Unlike NS algorithms, which look for a total coverage of the nonself space, DERA
searches for an appropriate distribution of eff ector and regulatory cells throughout
the space. By combining both regulatory and eff ector cells, to recognize normal
and abnormal operation, respectively, DERA's dynamic behavior mediated by cyto-
kines is able to indicate the severity of a fault. h e proposed approach was tested
on the DADAMICS fault-detection benchmark problem, and it was able to attain
considerably lower false-positives than other approaches, because regulatory cells
suppress the activation of eff ector cells.
7.6 Application to Scheduling
Creating optimal schedules in a constantly changing environment is not easy. h e
purpose of scheduling is to allocate a set of limited resources to tasks over time.
Ishida (1997) and Mori et al. (1994) proposed and developed an immune algorithm
that can create adaptive scheduling system based on the metaphors of somatic hyper-
mutation and immune network theory. Mori et al. (1994) built on this immune
algorithm by addressing the issue of batch sizes and combinations of sequence
orders, which optimized objective functions. In these works, antigens are consid-
ered as input data or disturbances in the optimization problem, and antibodies are
considered as possible schedules. Proliferation of the antibodies is controlled by an
immune network metaphor where stimulation and suppression are modeled in the
algorithm. h is assists in the control of antibody (or new solution) production. h e
T cell eff ect in this algorithm is ignored. h e authors claim that their algorithm
is an eff ective optimization algorithm for scheduling and was shown to be good
at fi nding optimal schedules. h e application of this algorithm to a dynamically
changing environment has been attempted by Mori et al. (1998). Here, the authors
considered antibodies as a single schedule and antigens to be possible changes to
the schedule. h eir system produced a set of antibodies (schedules) that can cover
the whole range of possible changes in the antigen set.
An AIS was utilized by Coello et al. (2003) to solve job-shop scheduling prob-
lems (JSSP) using clonal selection, hypermutation, and an antibody library to con-
struct solutions. h e purpose of JSSP is to fi nd an optimum schedule that gives the
minimum duration to complete all the jobs ( n jobs for m machines). It is an opti-
mization problem for particular objectives where certain criteria are met during the
assignment. A permutation representation (extensions of CLONALG) is adapted