You are on page 1of 4

MOEA Framework extension

The goal of our project is to extend MOEA (http://www.moeaframework.org), a Java


framework for working with multiobjective evolutionary algorithms. There are several
problems and metrics that we decided to implement. We chose problems that have
at least four objectives and are not included in the framework.
MOEA already includes implementations for the following metrics and problems:
Metrics:
Performance Metric
Hypervolume
Generational Distance
Inverted Generational Distance
Spacing
Additive-Indicator
Maximum Pareto Front Error
Native algorithms:
NSGA-2,3
epsilon-3
epsilon-MOEA
MOEA-D
GDE3
Random search
Native plugins for Borg

Problems:
Belegundu, Binh, Binh2, Binh3, Binh4, CF1, CF10, CF2, CF3, CF4, CF5, CF6, CF7, C
F8, CF9, DTLZ, Fonseca, Fonseca2, Jimenez, Kita,Kursawe, Laumanns, Lis, LZ, Mu
rata, Obayashi, OKA1, OKA2, Osyczka, Osyczka2, Poloni, ProblemStub, Quagliar
ella, Rendon, Rendon2,Schaffer, Schaffer2, Srinivas, Tamaki, Tanaka, UF1, UF10,
UF11, UF12, UF2, UF3, UF4, UF5, UF6, UF7, UF8, UF9, Viennet, Viennet2,Viennet
3, Viennet4, WFG, ZDT

Metrics to be implemented in MOEA framework


metric C
http://www.cs.ubbcluj.ro/~cgrosan/grosanCAIM.pdf
HVR metric:
http://books.google.ro/books?id=OSTn4GSy2uQC&pg=PA333&lpg=PA333&dq
=hvr+metric&source=bl&ots=tEgsqzMod8&sig=GpxXpxIvN4DmkPLi0YyHO442
M_4&hl=ro&sa=X&ei=W59TVOajMKXQ7AarloFQ&ved=0CBIQ6AEwAA#v=onep
age&q=hvr%20metric&f=false
capacity metrics: ONVG, C1, C2
Diversity: delta', m3, delta*
Convergence-diversity: Averaged Hausdorff Distance
http://www.ntu.edu.sg/home/zhangj/paper/siwei-measure.pdf

Test problems to be implemented in MOEA framework


(if not specified all objectives are to be minimized)

1. mQAP
Definition of the problem:

n is the number of locations, aij is the distance between location i and location j, bijk is
the k-th flow from facility i to facility j, i gives the location of facility i in
permutation from P(n).
Reference: http://www.cs.man.ac.uk/~jknowles/mQAP/
2. FES3
Definition of the problem:

Parameter domain: [0,1]


Reference: http://ro.ecu.edu.au/cgi/viewcontent.cgi?article=3021&context=ecuworks
(page 27)
3. ZLT1
Definition of the problem:

Parameter domain: [-1000,1000]


Reference: http://ro.ecu.edu.au/cgi/viewcontent.cgi?article=3021&context=ecuworks
(page 27)

4. Rectangle problem
Definition of the problem:
The Rectangle problem minimises the Euclidean distance from a solution,
= (x1, x2), to four given lines, which are, in pairs, parallel to the coordinate axes (o1 =
a1, o1 = a2, o2 = b1, and o2 = b1):

Reference:
http://www.brunel.ac.uk/~cspgmml1/Supplement/RectangleProblem(CEC2014).pdf
5. Uncertain multiobjective traveling salesman problem (UMTSP)
Definition of the problem:

Reference: http://www.sciencedirect.com/science/article/pii/S0377221714007334
6. Multiobjective Knapsack Problem (MKP)
Definition of the problem:
Given n items (i = 1, . . . , n) having m characteristics wij (j = 1, . . . , m)like weight,
volume, . . .and p profits cik (k = 1, . . . , p), some items should be selected to
maximize the p total profits while not exceeding the m knapsack capacities Wj
regarding the different characteristics.

where xi = 1 means that the item i is selected to be in the knapsack. It is assumed that
all coefficients cik, wij and Wj are nonnegative.
Reference: http://arxiv.org/ftp/arxiv/papers/1007/1007.4063.pdf

You might also like