Nelishia Pillay

• Phd (Computer Science)
• Professor at University of Pretoria

Neural architecture search (NAS) is a rapidly growing field which focuses on the automated design of neural network architectures. Genetic algorithms (GAs) have been predominantly used for evolving neural network architectures. Genetic programming (GP), a variation of GAs that work in the program space rather than a solution space, has not been as...

The authors would like to make the following corrections to this published paper [...]

The use of genetic algorithms (GAs) to evolve neural network (NN) weights has risen in popularity in recent years, particularly when used together with gradient descent as a mutation operator. However, crossover operators are often omitted from such GAs as they are seen as being highly destructive and detrimental to the performance of the GA. Desig...

Efficacy data from diverse chemical libraries, screened against the various stages of the malaria parasite Plasmodium falciparum, including asexual blood stage (ABS) parasites and transmissible gametocytes, serve as a valuable reservoir of information on the chemical space of compounds that are either active (or not) against the parasite. We postul...

Efficacy data from diverse chemical libraries, screened against the various stages of the malaria parasite Plasmodium falciparum , including asexual blood stage (ABS) parasites and transmissible gametocytes, serves as a valuable reservoir of information on the chemical space of compounds that are either active (or not) against the parasite. We post...

Facial expression recognition is an intriguing research area that has been explored and utilized in a wide range of applications such as health, security, and human-computer interactions. The ability to recognize facial expressions accurately is crucial for human-computer interactions. However, most of the facial expression analysis techniques have...

Search methodologies essentially explore a solution space to solve optimization problems. As the field has developed the effectiveness of exploring other spaces has been established. For example genetic programming explores the program space. Similarly, hyper-heuristics explore the heuristic space. In previous work the advantage of switching search...

Genetic programming tends to optimize complicated structures producing human-competitive results; therefore, it is applied to a wide range of problems such as classification and regression. This work experimentally performs a comparative study of Genetic programming variants, namely gene expression, grammatical evolution, Cartesian, multi-expressio...

The effectiveness of the Structure-Based Partial Solution Search (SBPSS) in solving the examination timetabling problem (ETP) was shown in previous work. The research presented in this paper extends this work by improving the previous version of the SBPSS. Two improvements were made, namely, additional search operators for exploitation and better c...

Selection hyper-heuristics have been used successfully to solve hard optimization problems. These techniques choose a heuristic or a group of heuristics to create a solution and/or improve it. In a prior study, we proposed an approach that changes the heuristic set from which the hyper-heuristic is allowed to choose dynamically and that led to impr...

Deep neural networks have proven to be effective in various domains, especially in natural language processing and image processing. However, one of the challenges associated with using deep neural networks includes the long design time and expertise needed to apply these neural networks to a particular domain. The research presented in this paper...

Maize yields worldwide are limited by foliar diseases that could be fungal, oomycete, bacterial, or viral in origin. Correct disease identification is critical for farmers to apply the correct control measures, such as fungicide sprays. Deep learning has the potential for automated disease classification from images of leaf symptoms. We aimed to de...

Research efforts in the improvement of artificial neural networks have provided significant enhancements in learning ability, either through manual improvement by researchers or through automated design by other artificial intelligence techniques, and largely focusing on the architecture of the neural networks or the weight update equations used to...

Barchan morphometric data have been used as proxies of meteorological and topographical data in environments where this data is lacking (such as other planetary bodies), gaining insights into barchan dune field dynamics such as barchan collision and sediment dynamics, and estimating migration speeds. However, manual extraction of this data is time-...

Research into the applicability of ant-based optimisation techniques for hyper-heuristics is largely limited. This paper expands upon the existing body of research by presenting a novel ant-based generation constructive hyper-heuristic and then investigates how different pheromone maps affect its performance. Previous work has focused on applying a...

Selection hyper-heuristics have proven to be effective in solving various real-world problems. Hyper-heuristics differ from traditional heuristic approaches in that they explore a heuristic space rather than a solution space. These techniques select constructive or perturbative heuristics to construct a solution or improve an existing solution resp...

Data classification is a real-world problem that is encountered daily in various problem domains. Genetic programming (GP) has proved to be one of the most versatile algorithms leading to its popularity as a classification algorithm. However, due to its large number of parameters, the manual design process of GP is considered to be a time consuming...

The task of generation constructive hyper-heuristics concerns itself with generating new heuristics for problem domains via some kind of mechanism that combines low-level heuristic components into new heuristics. The movie scene scheduling problem is a recently developed combinatorial problem for which there are relatively few low-level heuristics....

Usually, real-world time-series forecasting problems are dynamic. If such timeseries are characterized by mere concept shifts, a passive approach to learning become ideal to continuously adapt the model parameters whenever new data patterns arrive to cope with uncertainty in the presence of change. This work hybridizes a quantum-inspired particle s...

Multiple myeloma is a type of bone marrow cancer. Patient’s blood samples are analysed from protein gel strips and densitometer graphs which are then interpreted by a pathologist to diagnose multiple myeloma. This manual process of diagnosis is slow which is problematic as patients need to be diagnosed as soon as possible in order to prevent the co...

The paper provides a study of the use of hyper-heuristics on the movie scene scheduling problem. In particular, the paper extends the definition of the movie scene scheduling problem to include a new method of calculating the solution quality. The study is also a novel application of hyper-heuristics to the movie scene scheduling problem and demons...

The second sustainable development goal defined by the United Nations focuses on achieving food security and supporting sustainable agriculture. This paper focuses on one such initiative contributing to attaining this goal, namely, the identification or prediction of disease in crops. More specifically the paper examines the automated quantificatio...

Various machine learning techniques exist to perform regression on temporal data with concept drift occurring. However, there are numerous nonstationary environments where these techniques may fail to either track or detect the changes. This study develops a genetic programming-based predictive model for temporal data with a numerical target that t...

The special issue IEEE Transactions On Pattern Analysis And Machine Intelligence has a special section dedicated to AutoML. This special section is formed by 15 articles of outstanding quality that together comprise a snapshot of cutting edge AutoML research. The special section is a compilation of outstanding contributions on AutoML. The compilati...

Data classification provides effective solutions to various real-world problems in areas such as disease diagnosis, network intrusion detection, and financial forecasting, among others. Classification algorithms such as induction algorithms, e.g., ID3, and genetic programming are used to produce classifiers. The design of these classification algor...

The book has presented current trends and state-of-the-art advancements in the automated design of machine learning and search algorithms. In this context, we define automated design(AutoDes) to include automated algorithm/approach configuration, composition, and selection. This chapter provides a conclusion to the book by bringing together these c...

Optimisation has been with us since before the first humans opened their eyes to natural phenomena that inspire technological progress. Nowadays, it is quite hard to find a solver from the overpopulation of metaheuristics that properly deals with a given problem. This is even considered an additional problem. In this work, we propose a heuristic-ba...

Machine learning (ML) is an area of artificial intelligence that provides computer programmes with the capacity to autodidact and learn new skills from experience, without continued human programming. ML algorithms can analyse large data sets quickly and accurately, by supervised and unsupervised learning techniques, to provide classification and p...

This book presents recent advances in automated machine learning (AutoML) and automated algorithm design and indicates the future directions in this fast-developing area. Methods have been developed to automate the design of neural networks, heuristics and metaheuristics using techniques such as metaheuristics, statistical techniques, machine learn...

This paper extends a PSO-based nonlinear regression technique to dynamic environments whereby the induced model dynamically adjusts when an environmental change is detected. As such, this work hybridizes a PSO designed for dynamic environments with a least-squares approximation technique to induce structurally optimal nonlinear regression models. T...

Metaheuristics have become a widely used approach for solving a variety of practical problems. The literature is full of diverse metaheuristics based on outstanding ideas and with proven excellent capabilities. Nonetheless, oftentimes metaheuristics claim novelty when they are just recombining elements from other methods. Hence, the need for a stan...

Traditionally search techniques explore a single space to solve the problem at hand. This paper investigates performing search across more than one space which we refer to as bi-space search. As a proof of concept we illustrate this using the solution and heuristic spaces. In previous work two approaches for combining search across the heuristic an...

Limited attention has been paid to assessing the generality performance of hyper-heuristics. The performance of hyper-heuristics has been predominately assessed in terms of optimality which is not ideal as the aim of hyper-heuristics is not to be competitive with state of the art approaches but rather to raise the level of generality, i.e. the abil...

Search methodologies such as hyper-heuristics have been successfully used to automate the generation of perturbative heuristics to solve combinatorial optimization problems. However, the domain of automated generation of perturbative heuristics has generally not been well researched and very few works have actually been conducted in the area. In ad...

Since its inception genetic programming, and later variations such as grammar-based genetic programming and grammatical evolution, have contributed to various domains such as classification, image processing, search-based software engineering, amongst others. This paper examines the role that genetic programming has played in education. The paper f...

This paper defines a new combinatorial optimization problem, namely General Combinatorial Optimization Problem (GCOP), whose decision variables are a set of parametric algorithmic components, i.e. algorithm design decisions. The solutions of GCOP, i.e. compositions of algorithmic components, thus represent different generic search algorithms. Th...

Traditionally search algorithms have explored a solution space to solve optimization problems. However, as the field has advanced, in order to overcome the challenges posed by searching the solution space directly such as premature convergence, search has been applied to different spaces. These include genetic programming which explores the program...

Construction heuristics play an important role in solving combinatorial optimization problems. These heuristics are usually used to create an initial solution to the problem which is improved using optimization techniques such as metaheuristics. For examination timetabling and university course timetabling problems essentially graph colouring heuri...

Hybrid metaheuristics have proven to be effective at solving complex real-world problems. However, designing hybrid metaheuristics is extremely time consuming and requires expert knowledge of the different metaheuristics that are hybridized. In previous work, the effectiveness of automating the design of relay hybrid metaheuristics has been establi...

For the last decades, metaheuristics have become ever more popular as a tool to solve a large class of difficult optimization problems. However, determining the best configuration of a metaheuristic, which includes the program flow and the parameter settings, remains a difficult task. Adaptive metaheuristics (that change their configuration during...

The three articles in this special section focus on the development of automated design of machine learning and search algorithms. 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. Machine learning and search techniques pl...

This research builds on the hypothesis that the use of different fitness measures at the different generations of genetic programming (GP) is more effective than the convention of applying the same fitness measure individually throughout GP. Whereas the previous study used a genetic algorithm (GA) to induce the sequence in which fitness measures sh...

Genetic Programming (GP) is gaining increased attention as an effective method for inducing classifiers for data classification. However, the manual design of a genetic programming classification algorithm is a non-trivial time consuming process. This research investigates the hypothesis that automating the design of a GP classification algorithm f...

The previous chapters have introduced the four types of hyper-heuristics, presented the theoretical foundations and examined various applications of hyper-heuristics. This chapter provides an overview of some advanced topics and recent trends in hyper-heuristics, namely, hybrid hyper-heuristics, hyper-heuristics for automated design, automated desi...

Research into solving combinatorial optimization problems such as timetabling, vehicle routing and rostering problems has involved deriving techniques that improve the results obtained by existing techniques for known benchmark sets. These benchmark sets are made publicly available for performance comparisons of different techniques in solving thes...

Personnel scheduling problems arise from various real-world scenarios, including supermarket staff scheduling, call centre staff allocation, police force scheduling, and, the most studied, nurse rostering in hospitals. Due to the demands of quality healthcare, limited resources, and the tight constraints of specific legislation worldwide, the nurse...

In solving combinatorial optimization problems, a low-level constructive heuristic is used to create an initial solution, which forms a starting point for optimization techniques to solve the problem. These heuristics are problem dependent and are rules of thumb, manually derived based on human intuition. Deriving constructive heuristics is a time-...

Examination timetabling represents one of the earliest and most studied problem domains in hyper-heuristics (HH). Different interesting research issues have been addressed in the literature, from high-level heuristic selection mechanisms and acceptance criteria and designing intelligent low-level heuristics, to fundamental studies on the formal def...

Along with the continuous developments in hyper-heuristic (HH), various descriptive definitions for HH have emerged, leading to classifications of HH. Initially, hyper-heuristics have been defined as a search technique “to decide (select) at a higher abstraction level which low-level heuristics to apply” [51], “to combine simple heuristics” [162],...

Hyper-heuristics aim to provide heuristic algorithms of a higher level of generality that produce good results for all problems in a domain rather than just for one or two problem instances but poor results for the others. Cross-domain hyper-heuristics extend this scope of generality across domains. These hyper-heuristics aim at producing good resu...

Vehicle routing problems (VRP) [72, 50, 186] represent one of the most investigated combinatorial optimization problems [134], due to the problem complexity and their potential impact on real-world applications especially in logistics and supply chains. The basic VRP involves constructing a set of closed routes from and to a depot, each serviced by...

Selection constructive hyper-heuristics select a low-level heuristic at each point in the construction of a solution to a combinatorial optimization problem. As discussed in Chapter 1, the purpose of low-level construction heuristics is to construct complete solutions, or initial solutions for optimization. Solving a problem begins at an initial st...

Selection perturbative hyper-heuristics select which low-level perturbative heuristic to apply at each point of improvement to a given initial complete solution to a problem. The initial solution is usually created either randomly or using a constructive low-level heuristic. It is usually iteratively refined by applying a perturbative low-level heu...

The packing of items into a container or bin so as to minimize cost is a common problem experienced in industry. This chapter examines the use of hyperheuristics for solving bin packing problems presented in Appendix B.1. Selection constructive hyper-heuristics and generation constructive hyper-heuristics have been successfully employed to solve bi...

Recent research advances have been made in different types of hyper-heuristics (HH), namely selection HH and generation HH, employing both constructive and perturbative low-level heuristics (llh). Among the four types of HH, selection HH (Chapters 2, 3) received more research attention than generation HH (Chapters 4, 5). This may be due to the rese...

Low-level perturbative heuristics are used to improve a solution created either randomly or using a constructive heuristic for a combinatorial optimization problem. The low-level perturbative heuristics are problem dependent, and often move operators defined for the problem domain when solving the problem using local search techniques, e.g. the 2-o...