This document discusses issues with two published journal articles on the Teaching-Learning Based Optimization (TLBO) algorithm. It summarizes reviews from multiple reviewers on a paper that identified errors and inconsistencies in the TLBO papers and corresponding code. The reviewers found the paper convincing and recommended publishing it to alert readers to problems with the TLBO articles. When asked to check their findings against a new version of the TLBO code, the authors still found mismatches between the new and original codes.
1. AVN-21, Maribor, Dec 14, 2012
1/21
Komentarji na članek Gajanan
Waghmare: Comments on “A Note
on Teaching-Learning-Based
Optimization Algorithm”
Marjan Mernik
UNIVERSITY OF MARIBOR
FACULTY OF ELECTRICAL ENGINEERING
AND COMPUTER SCIENCE
2. AVN-21, Maribor, Dec 14, 2012
2/21
Komentarji na …
The first TLBO paper (from Computer-Aided design journal) was sent from Dr. Rao on 2/25/2011 to
the head of the first author’s laboratory, prof. Mernik.
Dr. Rao wrote in his e-mail:
Dear Sir,
Please find attached a research paper entitled "Teaching–learning-based optimization: A novel
method for constrained mechanical design optimization problems". This method works on the effect
of influence of a teacher on learners. Like other nature-inspired algorithms, this method is also a
population-based method and uses a population of solutions to proceed to the global solution. The
population is considered as a group of learners or a class of learners. Results show that this
algorithm is more effective and efficient than the other optimization methods for the mechanical
design optimization problems considered. This new optimization method can be easily extended to
other engineering design optimization problems.
I am sending this paper for your information and reference.
With sincere regards,
R. Venkata Rao
5. AVN-21, Maribor, Dec 14, 2012
5/21
Komentarji na …
On 3/3/2011 the TLBO Matlab code was sent to the head of the first author’s laboratory. Note, that
this TLBO code was not requested. The TLBO code was voluntarily sent.
Dr. Rao wrote on his e-mail message:
Dear Sir,
Please find atached the TLBO Code files. I am sending these files to some other researchers also
who have shown interest. I am going to improve this algorithm further.
With sincere regards,
R. Venkata Rao
DUPLICATE ELIMINATION PHASE WAS IDENTIFIED IN THE TLBO CODE !
8. AVN-21, Maribor, Dec 14, 2012
8/21
Komentarji na …
The second TLBO paper (from Information sciences journal) was sent from Dr. Rao on 8/31/2011
to the head of the first author’s laboratory.
Dear Sir,
Please find attached a research paper entitled "Teaching–Learning-Based Optimization: An
optimization method for continuous non-linear large scale problems". The proposed method is
based on the effect of the influence of a teacher on the output of learners in a class. The basic
philosophy of the method is explained step-by-step by means of Rastrigin function. The
effectiveness of the method is tested on many benchmark problems with different characteristics
and the results are compared with other population based methods.
I am sending this paper only for your information and academic reference. If you are interested in
the algorithm, please contact me so that I can send the MATLAB code.
With sincere regards,
R. Venkata Rao
10. AVN-21, Maribor, Dec 14, 2012
10/21
Komentarji na …
Reviewer #1: I have read the paper and try to control the matlab codes (I have also received
original codes from the author, Dr. RV Rao). It seems that the notes made by M Crepinsek,
S-H Liu and L Mernik are all correct. It is a pity to see that such wrong results are
published. It is really very hard to identify such mistakes from the submitted paper
unless we have the original codes. I think for such papers we need to ask for original
codes at least for the referees and editors.
Moreover, there is another problem with Rao et al's papers. Actually the paper published in
CAD journal and the paper in Information Sciences are very similar to each other. These
two papers were almost submitted at the same time period to these journals. One of them
published other one is in press. This can be considered self plagiarism or a slicing action.
Actually these two papers can be a single paper.
I strongly recommend urgent publication of the note of M Crepinsek, S-H Liu and L
Mernik. I also want to thank to them for their really trying but very important work.
Finally I recommend exclusion of Rao et al's paper from Information Sciences if it is
possible.
11. AVN-21, Maribor, Dec 14, 2012
11/21
Komentarji na …
Reviewer #2: The language used in the paper is very professional, however there are few very
minor typesetting errors. The paper very clearly expresses its case. Some very minor
misspelling errors that have been noticed should be corrected by the authors. The paper is very
well, precisely prepared and organized and does not need to go through the next
reviewing process.
Reviewer #3: The paper is well structured, and the experiments are detailed. The
inclusion of a reference to the source code is a good feature, although the reference may be a full
URL and not a compressed one. A important point of the paper is Section 4, the guidelines
will be very useful for researchers in general, in this matter the paper may be rewritten
as a guide of how to do a qualitative and quantitative analysis of new algorithms, and
give as a example the case of the Teaching-Learning Based Optimization Algorithm.
Reviewer #4: TBLO is a new algorithm whose characteristics are not deeply examined until now.
There should be more research and simulation done to see its extent and limits. However, what
are demonstrated in their study by the authors has proved that TBLO is a promising
new heuristic search method however its power should not be overestimated with
insufficient and wrong experimental setups.
12. AVN-21, Maribor, Dec 14, 2012
12/21
Komentarji na …
Reviewer #5: This paper reported several false statements and fundamental errors
that exist in two previous journal articles on the Teaching-Learning Based Optimisation
(TLBO) metaheuristics. The review process and experiments of this paper are coherent and
convincing. The foundings (and the guidelines) are very useful for researchers in the
related fields in many ways. I recommend the paper be accepted for the publication in
INS with minor corrections. In addition, it is recommended that the INS should adopt
measures to make sure that any potential readers of TLBO articles (at least the one
published by INS) by Rao et al. should be alerted with the key foundings from this
article.
Reviewer #6: I appreciate the authors for their Discussion which states so many
significant details of ethical issues in Research for Publication.
Reviewer #8: The detailed review report is attached.
AE: The authors are asked to compare with the actual codes provided by Reviewer #8.
13. AVN-21, Maribor, Dec 14, 2012
13/21
Komentarji na …
The third mismatch is in the manner in which the teacher and learning phases interact. We have
carefully studied the actual code provided by reviewer #8. Unfortunately, the provided code still
can’t be directly executed and tested since some functional implementations are missing. However,
the general algorithm’s inner workings can be easily identified. By comparing the code provided by
reviewer #8 (new TLBO code) and the code obtained earlier from TLBO inventors (old TLBO code),
we found that the new TLBO code has been changed and restructured.
In the new TLBO version each individual first undergoes the Teacher phase and after that
immediately the Learner phase follows. In the following iteration, the next individual is processed:
for i=1:length(Population)
... Teacher Phase ...
... Learner Phase ...
end
14. AVN-21, Maribor, Dec 14, 2012
14/21
Komentarji na …
Whilst in the old version of the TLBO code the whole population first undergoes the Teacher phase
and after that the Learner phase:
for i=1:size(pp,2)
... Teacher Phase ...
end
…
for i = 1 : 1 : length(Population)
... Learner Phase ...
end
15. AVN-21, Maribor, Dec 14, 2012
15/21
Komentarji na …
The fourth mismatch between the two versions of the TLBO algorithm is how elitism is
implemented. In the old TLBO code the following elitism was implemented:
Keep=2;
for GenIndex = 1 : OPTIONS.Maxgen
for i = 1 : Keep
chromKeep(i,:) = Population(i).chrom;
costKeep(i) = Population(i).cost;
end
... Teacher Phase ... ... Learner Phase ...
Population = PopSort(Population);
n = length(Population);
for i = 1 : Keep
Population(n-i+1).chrom = chromKeep(i,:);
Population(n-i+1).cost = costKeep(i);
end
Population = ClearDups(Population, MaxParValue, MinParValue);
Population = PopSort(Population);
end
16. AVN-21, Maribor, Dec 14, 2012
16/21
Komentarji na …
In the actual code provided by reviewer #8 the classic elitism has been replaced with the following
code, which can be regarded as a special elitist method:
for GenIndex = 1 : OPTIONS.Maxgen
... Teacher Phase ...
... Learner Phase ...
for i = 1 : length(Population)/2
ii=i+length(Population)/2-1;
if Population(ii).cost<Population(i).cost
Population(i).chrom =Population(ii).chrom;
Population(i).cost=Population(ii).cost;
end
end
...
Population = PopSort(Population);
Population = ClearDups(Population, MaxParValue, MinParValue);
end
18. AVN-21, Maribor, Dec 14, 2012
18/21
Komentarji na …
Response to AE after 2nd revision:
We are completely aware of the fact that reviewers have different opinions. In some cases even
diametrical. This is the reason why papers are not usually sent to only one or two reviewers. In the
case of diametrical opinion it would be hard to make a final decision. We have tried to delete as
many statements that were problematic for reviewer #8 as possible. See TLBO_R2_doc.pdf for all
changes in this round. However, many of the requests have no justification or are based on wrong
assumptions that results from experiments with different settings can be compared, and valid
conclusions can be drawn.
19. AVN-21, Maribor, Dec 14, 2012
19/21
Komentarji na …
Yet another letter from Dr. Rao on 12/03/2012
Dear Sir,
Please find six attachments related to the recently proposed "Teaching-Learning-Based
Optimization (TLBO) Algorithm". The proposed algorithm is an "algorithm-specific parameter-less"
one and can be used for solving the constrained and unconstrained optimization problems.
(1). Already the researchers have started applying this new algorithm to their research problems
and the first file gives you the details of research papers published so far using the proposed
algorithm.
(2). The second file is a research paper and presents the correct understanding on
TLBO algorithm. The paper was authored by Gajanan Waghmare and it corrects the
note published by Črepinšek et al. (2012) on TLBO in the journal of "Information
Sciences".
(3). The elitist version of TLBO algorithm and its applications to constrained and unconstrained
optimization problems is presented in files 3 and 4 respectively. The code of elitist TLBO is given in
the Appendices of these papers. Details on how to run the code are given in the Appendix of file 4..
(4). The attachments 5 and 6 are two sample papers describing the applications of TLBO algorithm.
The papers being sent to you are only for your academic reference.
…
With best regards,
R. Venkata Rao