Domain adaptation reuses the knowledge learned from an existing (source) domain to classify unlabeled data from another related (target) domain. However, the two domains have different data distributions. Common approaches to bridge the two distributions are selecting/reweighting instances, building domain-invariant feature subspaces, or directly building adaptive classifiers. Recent domain adaptation work has shown that combining the above first two approaches before applying the third approach achieves better performance than performing each approach individually. However, most existing instance selection approaches are based on a ranking mechanism, ignore interdependences between instances, and require a predefined number of selected instances. Furthermore, adaptive classifiers are sensitive to their parameters which are challenging to optimize due to the lack of target labeled instances. This article introduces a novel evolutionary instance selection approach for domain adaptation. We propose a compacted representation and an efficient fitness function for particle swarm optimization to automatically determine the number of selected instances while considering the interdependencies among instances. This article also proposes to use multiple partial classifiers to build a more reliable and robust adaptive classifier. The results show that evolutionary instance selection selects better instances than the ranking approach. In cooperation with multiple partial classifiers, the proposed algorithm achieves better performance than nine state-of-the-art and well-known domain adaptation approaches.

Neural Architecture Search (NAS) aims to automate the creation of Artificial Neural Networks, including Convolutional Neural Networks (CNN), lessening the reliance on labour-intensive manual design by human experts. A CNN architecture can be decomposed into a micro- and macro-architecture, each influenced by distinct design and optimisation strategies to con-tribute to the overall construction and performance of the CNN. Cellular Encoding (CE), an evolutionary computation technique, has been successfully used to represent diverse network topologies of varying complexities. Recently, CE has been applied to evolve CNN architectures, showing promising results. However, current CE-based NAS approaches focus on evolving either the micro-or macro-architectures without considering the evolution of both in the same algorithm. Evolving the micro- and macro-architecture together can increase the performance of evolved CNN architectures. This research introduces a novel two-stage hybrid approach, combining Genetic Algorithms (GA) and CE to evolve both the micro- and macro-architectures to synthesise CNNs for classification tasks. Candidate macro-architectures are evolved using a CE approach, while a GA approach is used to explore the micro-architecture search space. The proposed algorithm is evaluated across four commonly used datasets and compared against six NAS peer competitors and five state-of-the-art manually designed CNN architectures. The results validate the approach's high competitiveness, outperforming several peer competitors on image and text classification tasks.

Convolutional Neural Networks (CNNs) continue to achieve great success in classification tasks as innovative techniques and complex multi-path architecture topologies are introduced. Neural Architecture Search (NAS) aims to automate the design of these complex architectures, reducing the need for costly manual design work by human experts. Cellular Encoding (CE) is an evolutionary computation technique which excels in constructing novel multi-path topologies of varying complexity and has recently been applied with NAS to evolve CNN architectures for various classification tasks. However, existing CE approaches have severe limitations. They are restricted to only one domain, only partially implement the theme of CE, or only focus on the micro-architecture search space. This paper introduces a new CE representation and algorithm capable of evolving novel multi-path CNN architectures of varying depth, width, and complexity for image and text classification tasks. The algorithm explicitly focuses on the macro-architecture search space. Furthermore, by using a surrogate model approach, we show that the algorithm can evolve a performant CNN architecture in less than one GPU day, thereby allowing a sufficient number of experiment runs to be conducted to achieve scientific robustness. Experiment results show that the approach is highly competitive, defeating several state-of-the-art methods, and is generalisable to both the image and text domains.

The discounted knapsack problem (DKP) is an NP-hard combinatorial optimization problem that has gained much attention recently. Due to its high complexity, the usual solution combines a global search algorithm with a greedy local search algorithm to repair candidate solutions. The current greedy algorithms use a heuristic that ignores the items already in a candidate solution. This paper presents a new greedy algorithm for DKP that uses an expanded set of operators and better heuristics that are more effective at considering the selected items. Experimental results show that the proposed greedy algorithm has superior performance to three well-known greedy algorithms for DKP, both when operating independently and when combined with global search algorithms.

Particle swarm optimization (PSO) is a heuristic optimization algorithm generally applied to continuous domains. Binary PSO is a form of PSO applied to binary domains but uses the concepts of velocity and momentum from continuous PSO, which leads to its limited performance. In our previous work, we reformulated momentum as a stickiness property and velocity as a flipping probability to develop sticky binary PSO. The initial design provides a good base, but many key factors need to be investigated. In this article, we propose a new algorithm called dynamic sticky binary PSO by developing a dynamic parameter control strategy based on an investigation of exploration and exploitation in the binary search spaces. The proposed algorithm is compared with four state-of-the-art dynamic binary algorithms on two types of binary problems: 1) knapsack and 2) feature selection. The experimental results on the knapsack datasets show that the new velocity and momentum assist sticky binary PSO in evolving better solutions than the benchmark algorithms. On feature selection, the dynamic strategy takes the advantages of these two newly defined movement concepts to help the proposed algorithm to produce smaller feature subsets with higher classification performance. This is the first time in the binary PSO, the four important concepts, that is, velocity, momentum, exploration, and exploitation, are investigated systematically to capture the properties of the binary search spaces to evolve better solutions for binary problems.

Densely Connected Convolutional Networks (DenseNet) have demonstrated impressive performance on image classification tasks, but limited research has been conducted on using character-level DenseNet (char-DenseNet) architectures for text classification tasks. It is not clear what DenseNet architectures are optimal for text classification tasks. The iterative task of designing, training and testing of char-DenseNets is a time consuming task that requires expert domain knowledge. Evolutionary deep learning (EDL) has been used to automatically design CNN architectures for the image classification domain, thereby mitigating the need for expert domain knowledge. This study demonstrates the first work on using EDL to evolve char-DenseNet architectures for text classification tasks. A novel genetic programming-based algorithm (GP-Dense) coupled with an indirect-encoding scheme, facilitates the evolution of performant char-DenseNet architectures. The algorithm is evaluated on two popular text datasets, and the best-evolved models are benchmarked against four current state-of-the-art character-level CNN and DenseNet models. Results indicate that the algorithm evolves performant models for both datasets that outperform two of the state-of-the-art models in terms of model accuracy and three of the state-of-the-art models in terms of parameter size.

Domain adaptation utilizes learned knowledge from an existing domain (source domain) to improve the classification performance of another related, but not identical, domain (target domain). Most existing domain adaptation methods first perform domain alignment, then apply standard classification algorithms. Transfer classifier induction is an emerging domain adaptation approach that incorporates the domain alignment into the process of building an adaptive classifier instead of using a standard classifier. Although transfer classifier induction approaches have achieved promising performance, they are mainly gradient-based approaches which can be trapped at local optima. In this article, we propose a transfer classifier induction algorithm based on evolutionary computation to address the above limitation. Specifically, a novel representation of the transfer classifier is proposed which has much lower dimensionality than the standard representation in existing transfer classifier induction approaches. We also propose a hybrid process to optimize two essential objectives in domain adaptation: 1) the manifold consistency and 2) the domain difference. Particularly, the manifold consistency is used in the main fitness function of the evolutionary search to preserve the intrinsic manifold structure of the data. The domain difference is reduced via a gradient-based local search applied to the top individuals generated by the evolutionary search. The experimental results show that the proposed algorithm can achieve better performance than seven state-of-the-art traditional domain adaptation algorithms and four state-of-the-art deep domain adaptation algorithms.

Feature selection is an important task in machine learning that has two main objectives: 1) reducing dimensionality and 2) improving learning performance. Feature selection can be considered a multiobjective problem. However, it has its problematic characteristics, such as a highly discontinuous Pareto front, imbalance preferences, and partially conflicting objectives. These characteristics are not easy for existing evolutionary multiobjective optimization (EMO) algorithms. We propose a new decomposition approach with two mechanisms (static and dynamic) based on multiple reference points under the multiobjective evolutionary algorithm based on decomposition (MOEA/D) framework to address the above-mentioned difficulties of feature selection. The static mechanism alleviates the dependence of the decomposition on the Pareto front shape and the effect of the discontinuity. The dynamic one is able to detect regions in which the objectives are mostly conflicting, and allocates more computational resources to the detected regions. In comparison with other EMO algorithms on 12 different classification datasets, the proposed decomposition approach finds more diverse feature subsets with better performance in terms of hypervolume and inverted generational distance. The dynamic mechanism successfully identifies conflicting regions and further improves the approximation quality for the Pareto fronts.

Character-level convolutional neural networks (char-CNN) require no knowledge of the semantic or syntactic structure of the language they classify. This property simplifies its implementation but reduces its classification accuracy. Increasing the depth of char-CNN architectures does not result in breakthrough accuracy improvements. Research has not established which char-CNN architectures are optimal for text classification tasks. Manually designing and training char-CNNs is an iterative and time-consuming process that requires expert domain knowledge. Evolutionary deep learning (EDL) techniques, including surrogate-based versions, have demonstrated success in automatically searching for performant CNN architectures for image analysis tasks. Researchers have not applied EDL techniques to search the architecture space of char-CNNs for text classification tasks. This article demonstrates the first work in evolving char-CNN architectures using a novel EDL algorithm based on genetic programming, an indirect encoding and surrogate models, to search for performant char-CNN architectures automatically. The algorithm is evaluated on eight text classification datasets and benchmarked against five manually designed CNN architecture and one long short-term memory (LSTM) architecture. Experiment results indicate that the algorithm can evolve architectures that outperform the LSTM in terms of classification accuracy and five of the manually designed CNN architectures in terms of classification accuracy and parameter count.

Good and effective communication skills, in Global Competence Understanding global markets and business particular for technical and scientific writing in English, is practices: languages; work in multicultural important for engineering students who need to have good global environments; global ethics, safety and security; cultural, social and political whose first language is English. Generally, poor writing is also evident in the test statistics of the International English Language Testing System (IELTS) where writing has consistently the lowest score band among four test categories, even for those whom English is the first language. Intrigued by this phenomenon, we analyzed the writing of final year technical knowledge as well as competent soft skills. This is awareness increasingly being demanded by professional accreditation bodies. In response, institutions of higher education nowadays have courses or programs that teach students the soft skills part of these requirements, with the hope that by graduation, students can successfully demonstrate that they have met the accreditation standards. Anecdotal evidence from our academic staff, however, have shown that many graduating students have poor writing skills. This is also true for students engineering students, by examining their honors year capstone project reports. By examining the writing competence of these graduating students, the writing mistakes encountered in the reports were collected, analyzed and we propose to develop an online repository, together with the suggested corrections, so that students have online access to additional English writing support. Through this investigation, so-far, we have also uncovered that simple writing mistakes exist in the final version of an academic supervisor approved report. Some perspectives have been offered as to why this may happen.

The necessity of proposing algorithms that are effective in noisy image segmentation is clear in many real-world applications. This paper proposes a new algorithm for severely noisy image segmentation by looking at the proper choice of feature, and feature manipulation. We are using Discrete Wavelet Transformation (DWT) as a tool to provide our method with the proper feature, and then we manipulate it via wavelet shrinkage. Particle Swarm Optimization (PSO) is used to adaptively search for threshold values that produce the best segmentation results when applied in the wavelet shrinkage, and Fuzzy C-Means (FCM) is used as a fitness metric in PSO. The proposed method was tested on two different datasets being extremely contaminated with the common Gaussian noise. These tests indicate the superior performance and consistency of the proposed method in comparison to other state-of-the-art methods.

Semantic association computation is the process of quantifying the strength of a semantic connection between two textual units, based on different types of semantic relations. Semantic association computation is a key component of various applications belonging to a multitude of fields, such as computational linguistics, cognitive psychology, information retrieval and artificial intelligence. The field of semantic association computation has been studied for decades. The aim of this paper is to present a comprehensive survey of various approaches for computing semantic associations, categorized according to their underlying sources of background knowledge. Existing surveys on semantic computation have focused on a specific aspect of semantic associations, such as utilizing distributional semantics in association computation or types of spatial models of semantic associations. However, this paper has put a multitude of computational aspects and factors in one picture. This makes the article worth reading for those researchers who want to start off in the field of semantic associations computation. This paper introduces the fundamental elements of the association computation process, evaluation methodologies and pervasiveness of semantic measures in a variety of fields, relying on natural language semantics. Along the way, there is a detailed discussion on the main categories of background knowledge sources, classified as formal and informal knowledge sources, and the underlying design models, such as spatial, combinatorial and network models, that are used in the association computation process. The paper classifies existing approaches of semantic association computation into two broad categories, based on their utilization of background knowledge sources: knowledge-rich approaches; and knowledge-lean approaches. Each category is divided further into sub-categories, according to the type of underlying knowledge sources and design models of semantic association. A comparative analysis of strengths and limitations of various approaches belonging to each research stream is also presented. The paper concludes the survey by analyzing the pivotal factors that affect the performance of semantic association measures.

In classification, the task of domain adaptation is to learn a classifier to classify target data using unlabeled data from the target domain and labeled data from a related, but not identical, source domain. Transfer classifier induction is a common domain adaptation approach that learns an adaptive classifier directly rather than first adapting the source data. However, most existing transfer classifier induction algorithms are gradient-based, so they can easily get stuck at local optima. Moreover, they usually generate only a single classifier which might it the source data too well, which results in poor target accuracy. In this paper, we propose a population-based algorithm that can address the above two limitations. The proposed algorithm can re-initialize a population member to a promising region when the member is trapped at local optima. The population-based mechanism allows the proposed algorithm to output a set of classifiers which is more reliable than a single classifier. The experimental results show that the proposed algorithm achieves significantly better target accuracy than four state-of-the-art and well-known domain adaptation algorithms on three real-world domain adaptation problems.

Event detecting Wireless Sensor Networks (WSNs) with overlapping sensor nodes generate many redundant packets. Resource constraints in WSNs require minimizing the number of packets transmitted, so we want to eliminate all redundant packets. But this must be done carefully if we need to ensure reliable detection of all events happening across the network. Ensuring reliable transport of data is an important concern in resource-constrained WSNs. In a densely deployed sensor network, the sensing region of the sensor nodes often overlap with one another. Thus, data from sensor nodes that are in close proximity tend to be highly correlated. This introduces the concept of event reliability, where a reliable transfer of event data from each sensing region in a sensor network is sufficient. This paper proposes a distributed approach, i.e. Enhanced Event Reliability Protocol (EERP), that enables reliable transmission of event information to the sink node while minimizing redundant packets from nodes in close vicinity of one another. EERP builds on the spatio-temporal information of the originating nodes and employs an efficient combination of a greedy strategy with the use of multilateration algorithm in a unique way to perform event identification before reliably transmitting the event information towards the sink node. The performance of EERP is evaluated and compared against a base model, ERP and a commonly used scheme (SWIA). Our results show that EERP significantly improves event information delivery and energy efficiency while maintaining good coverage of events throughout the network.

Missing values are an unavoidable issue in many real-world datasets. One of the most popular approaches to classification with incomplete data is to use imputation to replace missing values with plausible values. However, powerful imputation methods are too computationally intensive when applying a classifier to a new unknown instance. This paper proposes new approaches to integrating imputation, clustering and feature selection for classification with incomplete data in order to improve efficiency without loss of accuracy. Clustering is used to reduce the number of instances used by the imputation. Feature selection is used to remove redundant and irrelevant features of training data which greatly reduces the cost of imputation. The paper also investigates the ability of Differential Evolution (DE) to search feature subsets with incomplete data. Results show that the integration of imputation, clustering and feature selection not only improves classification accuracy, but also dramatically reduces the computation time required to estimate missing values when classifying new instances.

Transfer learning aims to use acquired knowledge from existing (source) domains to improve learning performance on a different but similar (target) domains. Feature-based transfer learning builds a common feature space, which can minimize differences between source and target domains. However, most existing feature-based approaches usually build a common feature space with certain assumptions about the differences between domains. The number of common features needs to be predefined. In this work, we propose a new feature-based transfer learning method using particle swarm optimization (PSO), where a new fitness function is developed to guide PSO to automatically select a number of original features and shift source and target domains to be closer. Classification performance is used in the proposed fitness function to maintain the discriminative ability of selected features in both domains. The use of classification accuracy leads to a minimum number of model assumptions. The proposed algorithm is compared with four state-of-the-art feature-based transfer learning approaches on three well-known real-world problems. The results show that the proposed algorithm is able to extract less than half of the original features with better performance than using all features and outperforms the four benchmark semi-supervised and unsupervised algorithms. This is the first time Evolutionary Computation, especially PSO, is utilized to achieve feature selection for transfer learning.

Feature selection is an important but often expensive process, especially with a large number of instances. This problem can be addressed by using a small training set, i.e. a surrogate set. In this work, we propose to use a hierarchical clustering method to build various surrogate sets, which allows to analyze the effect of surrogate sets with different qualities and quantities on the feature subsets. Further, a dynamic surrogate model is proposed to automatically adjust surrogate sets for different datasets. Based on this idea, a feature selection system is developed using particle swarm optimization as the search mechanism. The experiments show that the hierarchical clustering method can build better surrogate sets to reduce the computational time, improve the feature selection performance, and alleviate overfitting. The dynamic method can automatically choose suitable surrogate sets to further improve the classification accuracy.

Many real-world datasets suffer from the unavoidable issue of missing values. Classification with incomplete data has to be carefully handled because inadequate treatment of missing values will cause large classification errors. Using imputation to transform incomplete data into complete data is a common approach to classification with incomplete data. However, simple imputation methods are often not accurate, and powerful imputation methods are usually computationally intensive. A recent approach to handling incomplete data constructs an ensemble of classifiers, each tailored to a known pattern of missing data. The main advantage of this approach is that it can classify new incomplete instances without requiring any imputation. This paper proposes an improvement on the ensemble approach by integrating imputation and genetic-based feature selection. The imputation creates higher quality training data. The feature selection reduces the number of missing patterns which increases the speed of classification, and greatly increases the fraction of new instances that can be classified by the ensemble. The results of experiments show that the proposed method is more accurate, and faster than previous common methods for classification with incomplete data.

Missing values are an unavoidable issue in many real-world datasets. Classification with incomplete data has to be addressed carefully because inadequate treatment often leads to a big classification error. Interval genetic programming (IGP) is an approach to directly use genetic programming to evolve an effective and efficient classifier for incomplete data. This paper proposes a method to improve IGP for classification with incomplete data by integrating IGP with ensemble learning to build a set of classifiers. Experimental results show that the integration of IGP and ensemble learning to evolve a set of classifiers for incomplete data can achieve better accuracy than IGP alone. The proposed method is also more accurate than other common methods for classification with incomplete data.

In illuminating tissues, a cylindrical diffuser (CD) has an advantage over regular laser sources due to its ability to illuminate a larger volume of the target tissue. This paper presents a co-registered large volume photoacoustic (PA) and ultrasonic (US) imaging for early endometrial cancer (EEC) detection using CD. It has the advantage that the US imaging system is outside the body and only the PA excitation device is inside the body, which makes the system more efficient and less invasive for EEC detection. The paper reports on two sets of experiments. The first set produced real-time PA images of blood vessel phantom. The second set demonstrated the imaging of pig uterus ex vivo. The results show that the system has the potential for imaging and characterizing of EEC.