IEEE Computational Intelligence Magazine - May 2018 - 16

Guest
Editorial

Nelishia Pillay
Department of Computer Science,
University of Pretoria, SOUTH AFRICA
Rong Qu
School of Computer Science,
University of Nottingham, UNITED KINGDOM
Dipti Srinivasan
Department of Electrical and Computer Engineering,
National University of Singapore, SINGAPORE
Barbara Hammer
CITEC Centre of Excellence, Bielefeld University,
GERMANY
Kenneth Sörensen
Department of Engineering Management,
University of Antwerp, BELGIUM

Automated Design of Machine Learning and Search Algorithms

achine learning and search
techniques play an important
role in solving real-world complex optimization problems in areas such
as transportation, data mining, computer
vision, computer security and software
development, amongst others. Given the
growing complexity of optimization
problems, the design of effective algorithms to solve these problems has become
more challenging and time consuming.
The design process is itself an optimization problem. Hence, there is a demand,
especially from industry and business, to
automate the design process, thereby to
remove the heavy reliance on human
experts and to reduce the man hours
involved in designing machine learning
and search algorithms.
Automated design includes parameter
tuning and control. Machine learning
and search algorithms require parameters
to be tuned, with the most appropriate
parameter values being problem dependent. For example, genetic operator
probabilities have to be decided for
genetic algorithms. Similarly, hyperparameters, e.g. the learning rate in deep
learning, needs to be chosen. There are
usually many options for hyperparameters. Automating the selection of these
Digital Object Identifier 10.1109/MCI.2018.2806988
Date of publication: 10 April 2018

There is a demand, especially from industry and
business, to automate the design of machine learning
and search algorithms, thereby removing the heavy
reliance on human experts.
values reduces the time and human
expertise required for this. Automating
parameter selection allows for the parameter values to be configured and adapted
dynamically during execution of the
algorithm, resulting in parameter-less
algorithms. For some algorithms, e.g.
evolutionary algorithms, it is also necessary to select which operators to use.
Automating this process allows for different operators to be applied at different
points in the algorithm.
Research into automated design of
machine learning and search algorithms
has also focused on generating new constructs used by these algorithms. These
constructs have ranged from construction
heuristics that are used to create initial
solutions that these algorithms optimize
further, creating operators used by search
algorithms, generating machine learning
workplans and architectures, to the
induction of solution algorithms and
software development.
Identifying an appropriate algorithm,
or combination of algorithms, to solve the

IEEE ComputatIonal IntEllIgEnCE magazInE | maY 2018

problem at hand has also been investigated
as part of automated design of machine
learning and search algorithms. One
approach to select an algorithm to solve a
particular problem is to identify a correlation between problem features and the
algorithm most suitable for solving the
problem. Rules representing a mapping of
this relationship have been induced for this
purpose. The second paper presented in
this special issue examines applying transformation functions to the problem features to overcome the problems of stagnation
and likeliness that arise in evolving and
applying such rules.
Hyper-heuristics, specifically selection hyper-heuristics, and evolutionary
algorithms have been examined for
hybridizing algorithms to solve problems. For example, the first paper in this
special issue reports on using a coevolutionary algorithm to combine support
vector machines to solve multiclass classification problems.
Various techniques have been examined for automated design, however the

Table of Contents for the Digital Edition of IEEE Computational Intelligence Magazine - May 2018