Iryna Yevseyeva

A Portfolio Optimization Approach to Selection in Multiobjective Evolutionary Algorithms

In this work, a new approach to selection in multiobjective evolutionary algorithms (MOEAs) is proposed. It is based on the portfolio selection problem, which is well known in financial management. The idea of optimizing a portfolio of investments according to both expected return and risk is transferred to evolutionary selection, and fitness assignment is reinterpreted as the allocation of capital to the individuals in the population, while taking into account both individual quality and population diversity. The resulting selection procedure, which unifies parental and environmental selection, is instantiated by defining a suitable notion of (random) return for multiobjective optimization. Preliminary experiments on multiobjective multidimensional knapsack problem instances show that such a procedure is able to preserve diversity while promoting convergence towards the Pareto-optimal front.

Multiobjective optimization of classifiers by means of 3D convex-hull-based evolutionary algorithms

The receiver operating characteristic (ROC) and detection error tradeoff (DET) curves are frequently used in the machine learning community to analyze the performance of binary classifiers. Recently, the convex-hull-based multiobjective genetic programming algorithm was proposed and successfully applied to maximize the convex hull area for binary classification problems by minimizing false positive rate and maximizing true positive rate at the same time using indicator-based evolutionary algorithms. The area under the ROC curve was used for the performance assessment and to guide the search. Here we extend this research and propose two major advancements: Firstly we formulate the algorithm in detection error tradeoff space, minimizing false positives and false negatives, with the advantage that misclassification cost tradeoff can be assessed directly. Secondly, we add complexity as an objective function, which gives rise to a 3D objective space (as opposed to a 2D previous ROC space). A domain specific performance indicator for 3D Pareto front approximations, the volume above DET surface, is introduced, and used to guide the indicator-based evolutionary algorithm to find optimal approximation sets. We assess the performance of the new algorithm on designed theoretical problems with different geometries of Pareto fronts and DET surfaces, and two application-oriented benchmarks: (1) Designing spam filters with low numbers of false rejects, false accepts, and low computational cost using rule ensembles, and (2) finding sparse neural networks for binary classification of test data from the UCI machine learning benchmark. The results show a high performance of the new algorithm as compared to conventional methods for multicriteria optimization.

A spam filtering multi-objective optimization study covering parsimony maximization and three-way classification

Display Omitted Advances on applications of multi-objective optimization to anti-SPAM filtering.Parsimony maximization of rule-based SPAM classifiers.Three-way classification balancing user effort and confidence level.Indicator-based/machine learning/decomposition-based evolutionary optimization. Classifier performance optimization in machine learning can be stated as a multi-objective optimization problem. In this context, recent works have shown the utility of simple evolutionary multi-objective algorithms (NSGA-II, SPEA2) to conveniently optimize the global performance of different anti-spam filters. The present work extends existing contributions in the spam filtering domain by using three novel indicator-based (SMS-EMOA, CH-EMOA) and decomposition-based (MOEA/D) evolutionary multi-objective algorithms. The proposed approaches are used to optimize the performance of a heterogeneous ensemble of classifiers into two different but complementary scenarios: parsimony maximization and e-mail classification under low confidence level. Experimental results using a publicly available standard corpus allowed us to identify interesting conclusions regarding both the utility of rule-based classification filters and the appropriateness of a three-way classification system in the spam filtering domain.

Nudging for Quantitative Access Control Systems

On the one hand, an access control mechanism must make a conclusive decision for a given access request. On the other hand, such a mechanism usually relies on one or several decision making processes, which can return partial decisions, inconclusive ones, or conflicting ones. In some cases, this information might not be sufficient to automatically make a conclusive decision, and the access control mechanism might have to involve a human expert to make the final decision. In this paper, we formalise these decision making processes as quantitative access control systems, which associate each decision with a measure, indicating for instance the level of confidence of the system in the decision. We then propose to explore how nudging, i.e., how modifying the context of the decision making process for that human expert, can be used in this context. We thus formalise when such a delegation is required, when nudging is applicable, and illustrate some examples from the MINDSPACE framework in the context of access control.

Survey on Anti-spam Single and Multi-objective Optimization

In this paper anti-spam filtering is presented as a cumbersome service, as opposing to a software product perspective. The human effort for setting up, adaptation, maintenance and tuning of filters for spam detection is stressed. Because choosing the proper scores (relevance) for the spam filters is essential to the accuracy of the anti-spam system and one of the biggest challenges for the Apache SpamAssassin project (the most widely adopted anti-spam open-source software), we present a survey on single and multi-objective optimization studies for this purpose. Our survey constitutes a contribution and a stimulus for further research on this open research topic, with particular emphasis on evolutionary multi-objective approaches.

Building and Using an Ontology of Preference-Based Multiobjective Evolutionary Algorithms

Integrating user preferences in Evolutionary Multiobjective Optimization EMO is currently a prevalent research topic. There is a large variety of preference handling methods originated from Multicriteria decision making, MCDM and EMO methods, which have been combined in various ways. This paper proposes a Web Ontology Language OWL ontology to model and systematize the knowledge of preference-based multiobjective evolutionary algorithms PMOEAs. Detailed procedure is given on how to build and use the ontology with the help of Protege. Different use-cases, including training new learners, querying and reasoning are exemplified and show remarkable benefit for both EMO and MCDM communities.

A Formal Model for Soft Enforcement: Influencing the Decision-Maker

We propose in this paper a formal model for soft enforcement, where a decision-maker is influenced towards a decision, rather than forced to select that decision. This novel type of enforcement is particularly useful when the policy enforcer cannot fully control the environment of the decision-maker, as we illustrate in the context of attribute-based access control, by limiting the control over attributes. We also show that soft enforcement can improve the security of the system when the influencer is uncertain about the environment, and when neither forcing the decision-maker nor leaving them make their own selection is optimal. We define the general notion of optimal influencing policy, that takes into account both the control of the influencer and the uncertainty in the system.

A Decision Making Model of Influencing Behavior in Information Security

Information security decisions typically involve a trade-off between security and productivity. In practical settings, it is often the human user who is best positioned to make this trade-off decision, or in fact has a right to make its own decision (such as in the case of ‘bring your own device’), although it may be responsibility of a company security manager to influence employees choices. One of the practical ways to model human decision making is with multi-criteria decision analysis, which we use here for modeling security choices. The proposed decision making model facilitates quantitative analysis of influencing information security behavior by capturing the criteria affecting the choice and their importance to the decision maker.Within this model, we will characterize the optimal modification of the criteria values, taking into account that not all criteria can be changed. We show how subtle defaults influence the choice of the decision maker and calculate their impact. We apply our model to derive optimal policies for the case study of a public Wi-Fi network selection, in which the graphical user interface aims to influence the user to a particular security behavior.

Quadcriteria optimization of binary classifiers: error rates, coverage, and complexity

This paper presents a 4-objective evolutionary multiobjective optimization study for optimizing the error rates (false positives, false negatives), reliability, and complexity of binary classifiers. The example taken is the email anti-spam filtering problem.

3D fast convex-hull-based evolutionary multiobjective optimization algorithm

Abstract The receiver operating characteristic (ROC) and detection error tradeoff (DET) curves have been widely used in the machine learning community to analyze the performance of classifiers. The area (or volume) under the convex hull has been used as a scalar indicator for the performance of a set of classifiers in ROC and DET space. Recently, 3D convex-hull-based evolutionary multiobjective optimization algorithm (3DCH-EMOA) has been proposed to maximize the volume of convex hull for binary classification combined with parsimony and three-way classification problems. However, 3DCH-EMOA revealed high consumption of computational resources due to redundant convex hull calculations and a frequent execution of nondominated sorting. In this paper, we introduce incremental convex hull calculation and a fast replacement for non-dominated sorting. While achieving the same high quality results, the computational effort of 3DCH-EMOA can be reduced by orders of magnitude. The average time complexity of 3DCH-EMOA in each generation is reduced from O ( n 2 log n ) to O ( n log n ) per iteration, where n is the population size. Six test function problems are used to test the performance of the newly proposed method, and the algorithms are compared to several state-of-the-art algorithms, including NSGA-III, RVEA, etc., which were not compared to 3DCH-EMOA before. Experimental results show that the new version of the algorithm (3DFCH-EMOA) can speed up 3DCH-EMOA for about 30 times for a typical population size of 300 without reducing the performance of the method. Besides, the proposed algorithm is applied for neural networks pruning, and several UCI datasets are used to test the performance.