Wei Pang | University of Aberdeen (original) (raw)

Papers by Wei Pang

2017 4th International Conference on Systems and Informatics (ICSAI), 2017

Gravitation field algorithm (GFA) is a novel optimization algorithm derived from the Solar Nebula... more Gravitation field algorithm (GFA) is a novel optimization algorithm derived from the Solar Nebular Disk Model (SNDM) in astronomy, based on the formation of planets, in recent years. In this research, an improved GFA with Optimal Detection (GFA-OD) is proposed for unconstrained optimization problems. Optimal Detection can efficiently locate the space that more likely contains the optimal solution(s) by initializing part of dust population randomly in the search space of a given problem, and then improves the accuracy of solutions. The comparison of results on four classical unconstrained optimization problems with varying dimensions demonstrates that the proposed GFA-OD outperforms many other classical heuristic optimization algorithms in accuracy, efficiency and running time in lower dimensions, such as Genetic Algorithm (GA) and Particle Swarm Optimization (PSO).

2019 18th IEEE International Conference On Machine Learning And Applications (ICMLA), 2019

Full bibliographic details must be given when referring to, or quoting from full items including ... more Full bibliographic details must be given when referring to, or quoting from full items including the author's name, the title of the work, publication details where relevant (place, publisher, date), pagination, and for theses or dissertations the awarding institution, the degree type awarded, and the date of the award.

Applied Soft Computing, 2020

Feature selection has been widely used in classification for improving classification accuracy an... more Feature selection has been widely used in classification for improving classification accuracy and reducing computational complexity. Recently, evolutionary computation (EC) has become an important approach for solving feature selection problems. However, firstly, as the datasets processed by classifiers become

Computational Materials Science, 2019

Self-healing materials with an inherent repair mechanism have been widely studied. However, the s... more Self-healing materials with an inherent repair mechanism have been widely studied. However, the self-healing efficiencies of most materials can only be measured by laboratory-based experiments, which can be time consuming and expensive. Inspired by modern machine learning approaches, we are interested in predicting the selfhealing efficiency of new bio-hybrid materials, as part of our ongoing EPSRC funded "Manufacturing Immortality" project. By modelling existing experimental data, predictive models can be built to forecast selfhealing efficiency. This has the potential to reduce the time input required by laboratory experiments, guide material and component selection, and inform hypotheses, thereby facilitating the design of novel self-healing materials. In this position paper, we first present preliminary knowledge and quantitative definitions of the selfhealing efficiency of materials. We then demonstrate several widely used machine learning approaches and review an experimental case of predictive modelling based on neural networks. Furthermore, and aiming to expedite self-healing material development, we propose an on-line ensemble learning framework as the whole system model for the optimization of predictive computational models. Finally, the rationality of our on-line ensemble learning framework is experimentally studied and validated.

Lecture Notes in Computer Science, 2014

Semi-supervised clustering on information networks combines both the labeled and unlabeled data s... more Semi-supervised clustering on information networks combines both the labeled and unlabeled data sets with an aim to improve the clustering performance. However, the existing semi-supervised clustering methods are all designed for homogeneous networks and do not deal with heterogeneous ones. In this work, we propose a semi-supervised clustering approach to analyze heterogeneous information networks, which include multi-typed objects and links and may contain more useful semantic information. The major challenge in the clustering task here is how to handle multi-relations and diverse semantic meanings in heterogeneous networks. In order to deal with this challenge, we introduce the concept of relation-path to measure the similarity between two data objects of the same type. Thereafter, we make use of the labeled information to extract different weights for all relation-paths. Finally, we propose SemiRPClus, a complete framework for semi-supervised learning in heterogeneous networks. Experimental results demonstrate the distinct advantages in effectiveness and efficiency of our framework in comparison with the baseline and some state-of-the-art approaches.

Journal of Computational Science, 2014

In this paper, a novel qualitative differential equation model learning (QML) framework named QML... more In this paper, a novel qualitative differential equation model learning (QML) framework named QML-Morven is presented. QML-Morven employs both symbolic and evolutionary approaches as its learning strategies to deal with models of different complexity. Based on this framework, a series of experiments were designed and carried out to: (1) investigate factors that influence the learning precision and minimum data requirement for successful learning; (2) address the scalability issue of QML systems.

Neurocomputing, 2012

Canonical correlations analysis (CCA) is often used for feature extraction and dimensionality red... more Canonical correlations analysis (CCA) is often used for feature extraction and dimensionality reduction. However, the image vectorization in CCA breaks the spatial structure of the original image, and the excessive dimension of vector often brings the curse of dimensionality problem. In this paper, we propose a novel feature extraction method based on CCA in multi-linear discriminant subspace by encoding an action sample as a high-order tensor. An optimization approach is presented to iteratively learn the discriminant subspace by unfolding the tensor along different tensor modes. It retains most of the underlying data structure including the spatio-temporal information, and alleviates the curse of dimensionality problem. At the same time, an incremental scheme is developed for multi-linear subspace online learning, which can improve the discriminative capability efficiently and effectively. The nearest neighbor classifier (NNC) is exploited for action classification. Experiments on Weizmann database showed that the proposed method outperforms the state-of-the-art methods in terms of accuracy. The proposed method has low time complexity and is robust against partial occlusion.

Journal of Applied Mathematics, 2014

In this paper, we proposed a new building recognition method named subregion’s multiscale gist fe... more In this paper, we proposed a new building recognition method named subregion’s multiscale gist feature (SM-gist) extraction and corresponding columns information based dimensionality reduction (CCI-DR). Our proposed building recognition method is presented as a two-stage model: in the first stage, a building image is divided into 4 × 5 subregions, and gist vectors are extracted from these regions individually. Then, we combine these gist vectors into a matrix with relatively high dimensions. In the second stage, we proposed CCI-DR to project the high dimensional manifold matrix to low dimensional subspace. Compared with the previous building recognition method the advantages of our proposed method are that (1) gist features extracted by SM-gist have the ability to adapt to nonuniform illumination and that (2) CCI-DR can address the limitation of traditional dimensionality reduction methods, which convert gist matrices into vectors and thus mix the corresponding gist vectors from dif...

Advances in Intelligent Systems and Computing, 2019

In this paper we propose DeepSwarm, a novel neural architecture search (NAS) method based on Swar... more In this paper we propose DeepSwarm, a novel neural architecture search (NAS) method based on Swarm Intelligence principles. At its core DeepSwarm uses Ant Colony Optimization (ACO) to generate ant population which uses the pheromone information to collectively search for the best neural architecture. Furthermore, by using local and global pheromone update rules our method ensures the balance between exploitation and exploration. On top of this, to make our method more efficient we combine progressive neural architecture search with weight reusability. Furthermore, due to the nature of ACO our method can incorporate heuristic information which can further speed up the search process. After systematic and extensive evaluation, we discover that on three different datasets (MNIST, Fashion-MNIST, and CIFAR-10) when compared to existing systems our proposed method demonstrates competitive performance. Finally, we open source DeepSwarm 1 as a NAS library and hope it can be used by more deep learning researchers and practitioners.

Artificial Intelligence Review

In recent years, research on Physarum polycephalum has become more popular after Nakagaki (AIR 40... more In recent years, research on Physarum polycephalum has become more popular after Nakagaki (AIR 407: 6803-470, 2000) performed their famous experiment showing that Physarum was able to find the shortest route through a maze. Subsequent researches have confirmed the ability of Physarum-inspired algorithms to solve a wide range of real-world applications. In contrast to previous reviews that either focus on biological aspects or bio-inspired applications, here we present a comprehensive review that highlights recent Physarum polycephalum biological aspects, mathematical models, and Physarum bio-inspired algorithms and their applications. The novelty of this review stems from our exploration of Physarum intelligent behaviour in competition settings. Further, we have presented our new model to simulate Physarum in competition, where multiple Physarum interact with each other and with their environments. The bio-inspired Physarum in competition algorithms proved to have great potentials f...

ArXiv, 2019

Current Neural Architecture Search techniques can suffer from a few shortcomings, including high ... more Current Neural Architecture Search techniques can suffer from a few shortcomings, including high computational cost, excessive bias from the search space, conceptual complexity or uncertain empirical benefits over random search. In this paper, we present ImmuNeCS, an attempt at addressing these issues with a method that offers a simple, flexible, and efficient way of building deep learning models automatically, and we demonstrate its effectiveness in the context of convolutional neural networks. Instead of searching for the 1-best architecture for a given task, we focus on building a population of neural networks that are then ensembled into a neural network committee, an approach we dub 'Neural Committee Search'. To ensure sufficient performance from the committee, our search algorithm is based on an artificial immune system that balances individual performance with population diversity. This allows us to stop the search when accuracy starts to plateau, and to bridge the pe...

Memetic Computing, 2020

Inferring dynamic system models from observed time course data is very challenging compared to st... more Inferring dynamic system models from observed time course data is very challenging compared to static system identification tasks. Dynamic system models are complicated to infer due to the underlying large search space and high computational cost for simulation and verification. In this research we aim to infer both the structure and parameters of a dynamic system simultaneously by particle swarm optimization (PSO) improved by efficient stratified sampling approaches. More specifically, we enhance PSO with two modern stratified sampling techniques, i.e., Latin hyper cube sampling (LHS) and Latin hyper cube multi dimensional uniformity (LHSMDU). We propose and evaluate two novel swarm-inspired algorithms, LHS-PSO and LHSMDU-PSO, which can be used particularly to learn the model structure and parameters of complex dynamic systems efficiently. The performance of LHS-PSO and LHSMDU-PSO is further compared with the original PSO and genetic algorithm (GA). We chose real-world cancer biolo...

Proceedings of the 9th annual conference companion on Genetic and evolutionary computation, 2007

In this paper, a modified Clonal Selection Algorithm (CSA) is proposed to learn qualitative compa... more In this paper, a modified Clonal Selection Algorithm (CSA) is proposed to learn qualitative compartmental models. Different from traditional AI search algorithm, this populationbased approach employs antibody repertoire to perform random search, which is suitable for the ragged and multi-modal landscape of qualitative model space. Experimental result shows that this algorithm can obtain the same kernel sets and learning reliability as previous work for learning the twocompartment model, and it can also search out the target model when learning the more complex three-compartment model. Although this algorithm does not succeed in learning the four-compartment model, promising result is still obtained.

In this paper, the learning of qualitative twocompartment metabolic models is studied under the c... more In this paper, the learning of qualitative twocompartment metabolic models is studied under the conditions of different types and numbers of hidden variables. For each condition, all the experiments, each of which takes one of the subsets of the complete qualitative states as training data, are tested one by one. In order to conduct the experiments more efficiently, a backtracking algorithm with forward checking is introduced to search out all the well-posed qualitative models as candidate solutions. Then these candidate solutions are verified by a fuzzy qualitative engine JMorven to find the target models. Finally the learning reliability and kernel set under different conditions is calculated and analyzed.

Physarum polycephalum (Physarum for short) is an example of plasmodial slime moulds that are clas... more Physarum polycephalum (Physarum for short) is an example of plasmodial slime moulds that are classified as a fungus "Myxomycetes". In recent years, research on Physarum has become more popular after Nakagaki et al. (2000) performed his famous experiments showing that Physarum was able to find the shortest route through a maze. Physarum) may not have a central information processing unit like a brain, however, recent research has confirmed the ability of Physarum-inspired algorithms to solve a wide range of NP-hard problems. This review will through light on recent Physarum polycephalum biological aspects, mathematical models, and Physarum bio-inspired algorithms and its applications. Further, we have added presented our new model to simulate Physarum in competition, where multiple Physarum interact with each other and with their environments. The bio-inspired Physarum in competition algorithms proved to have great potentials in dealing with graph-optimisation problems in a...

arXiv (Cornell University), Jan 22, 2021

2020 IEEE Congress on Evolutionary Computation (CEC), 2020

This paper addresses two practical problems: the classification and prediction of properties for ... more This paper addresses two practical problems: the classification and prediction of properties for polymer and glass materials, as a case study of evolutionary learning for tackling soft margin problems. The presented classifier is modelled by support vectors as well as various kernel functions, with its hard restrictions relaxed by slack variables to be soft restrictions in order to achieve higher performance. We have compared evolutionary learning with traditional gradient methods on standard, dual and soft margin support vector machines, built by polynomial, Gaussian, and ANOVA kernels. Experimental results for data on 434 polymers and 1,441 glasses show that both gradient and evolutionary learning approaches have their advantages. We show that within this domain the chosen gradient methodology is beneficial for standard linear classification problems, whilst the evolutionary methodology is more effective in addressing highly non-linear and complex problems, such as the soft margin problem.

ArXiv, 2020

In this research, we propose ImmuNetNAS, a novel Neural Architecture Search (NAS) approach inspir... more In this research, we propose ImmuNetNAS, a novel Neural Architecture Search (NAS) approach inspired by the immune network theory. The core of ImmuNetNAS is built on the original immune network algorithm, which iteratively updates the population through hypermutation and selection, and eliminates the self-generation individuals that do not meet the requirements through comparing antibody affinity and inter-specific similarity. In addition, in order to facilitate the mutation operation, we propose a novel two-component based neural structure coding strategy. Furthermore, an improved mutation strategy based on Standard Genetic Algorithm (SGA) was proposed according to this encoding method. Finally, based on the proposed two-component based coding method, a new antibody affinity calculation method was developed to screen suitable neural architectures. Systematic evaluations demonstrate that our system has achieved good performance on both the MNIST and CIFAR-10 datasets. We open-source ...

Online and virtual spaces comprise a myriad of ad-hoc networks and online communities. Such commu... more Online and virtual spaces comprise a myriad of ad-hoc networks and online communities. Such communities are composed of smart devices, agents, systems and people who seek to interact in one way or another. We argue that the task of detecting anomalies in such settings is non-trivial. The complexity is further compounded since there is no clear cut definition/specification of what normal behaviour is, and how far out an outlier should be before it is detected as an anomaly. This is often the case with online and virtual spaces as there is little or no regulation of the interactions between the various players in online communities. Hence, detecting anomalous behaviour in such settings poses a huge challenge. In this paper, we investigate how evolutionary clustering could be exploited to support decision makers, designers and data scientists in the autonomous detection of anomalies in online and virtual spaces. We present preliminary ideas in tackling this issue using a freeform onlin...