Automatic Teaching–Learning-Based Optimization: A Novel Clustering Method for Gene Functional Enrichments - Computational Intelligence Techniques for Comparative Genomics - page 20

TLBO relevance to cluster analysis was shown in 2012 by Amiri [ 12 ]; this study
was accomplished by testing on quite a few numbers of datasets. Automatic clus-
tering in multi-objective optimization framework using differential evolution was
shown by Suresh et al. in 2011 [ 13 ]. The experimental result over different datasets
proves the variations of DE that are desired for doing automatic clustering. Auto-
matic clustering using genetic algorithms and generating optimality with Pareto
front is well demonstrated by the same set of authors Suresh et al. in 2009 [ 14 ].
Cluster evaluation, ranking, and validation using CVIs are effectively shown by Liu
et al. in 2005 [ 15 ]. The conceptualization toward

tting in automatic clustering into

TLBO in the paper was from [ 13 , 14 ].
Satpathy et al. in 2013 [ 16 ] brought an improved version of TLBO by using
orthogonal design. This change was proved as a statistically effect method to
generate an optimal offspring in EA. In the recent past, automatic clustering in
TLBO was shown by Naik et al. in 2012 [ 17 ] using fuzzy c means. The results were
well demonstrated over arti cial and real datasets. In 2014, Murthy et al. [ 18 ] used

automatic clustering in TLBO to nd optimal number of clusters and shown

potential results proving the ef ciency of algorithm.

The proposal toward using automatic clustering in TLBO over microarray
datasets was from the article published by Suresh et al. in 2009 [ 19 ] and Pavan et al.
in 2011 [ 20 ]. Both these articles use a test suite to compare results over the gene
datasets. The acquired optimal numbers of clusters are veried by using CVIs.

3 TLBO

TLBO algorithm is a teaching learning methodology-motivated population-based

algorithm, proposed by Rao et al. [ 1

-

4 , 10 , 11 ] which focused around the impact of
a teacher on the after effect of learners in a class. In this optimization algorithm, the
faction of learners are assumed as population and diverged con

-

guration of vari-
ables are treated as distinctive subjects accessible to the learners, and their result is
comparable to the

tness estimation value of this optimization issue. In the whole
population, the best solution is treated as the teacher.
Teacher phase: It is included as the rst segment of TLBO, where learners gain

knowledge from the teacher. In this phase, the teacher attempts to increase the mean
value of the class room from any value mean 1 to his or her echelon I A . But sensibly
it is not promising and a teacher can move the mean of the class room mean 1 to any
other value mean 2 which is healthier than mean 1 depending on his or her compe-
tence. Considered mean j be the mean and I i be the teacher at any iteration i. Now,
teacher I i will try to improve the existing mean mean j toward it so the new mean
will be I i designated as mean new , and the difference between the existing mean and
new mean is given as