Search

Adaptive modulation and coding schemes play a crucial role in ensuring robust data transfer in wireless communications, especially when faced with changes or interference in the transmission channel. These schemes involve the use of variable coding rates, which can be achieved normally through code puncturing or shortening, and have been adopted in 4G and 5G communication standards. In recent works, auto-encoders for wireless communications have demonstrated the ability to learn short code representations that achieve gains over conventional codes. Such a methodology is attractive as it can learn optimal representations under a variety of channel conditions. However, due to its structure the auto-encoder does not currently support multiple code rates with a single model. This article draws upon the discipline of multi-task learning, as it applies to deep learning and therefore devises a branching architecture for the auto-encoder and custom training algorithm in training transmitter and receiver for adaptive modulation and coding. In this article we aim to demonstrate improvements in Block Error Rate over conventional methods in the Additive White Gaussian Noise channel, and to analyse the performance of the model under Rayleigh fading channels without retraining the auto-encoder on the new channel. This article demonstrates a novel approach towards training auto-encoder models to jointly learn adaptive modulation and coding schemes framed as a multi-task learning problem. The research outcomes extend end-to-end learning approaches to the design of adaptive wireless communications systems. [Display omitted] • Proposes an end-to-end machine learning architecture enabling multiple code rates. • Custom training/multi-task learning produces competitive error-rate performance. • Proposed approach outperforms several traditional coding techniques for short codes. • Proposed approach is versatile in adapting to varying channels without retraining. • Demonstrates machine learning design methods for resilient wireless communications. [ABSTRACT FROM AUTHOR]

Decision Trees (DTs) are a class of supervised learning models that are widely used for both classification and regression applications. They are well-known for their interpretability and robustness, which have led them to remain popular even 60 years after they were first proposed. However, because traditional tree algorithms use greedy methods that are prone to suboptimality, several works have explored the usage of evolutionary algorithms instead. Although these algorithms are often reported to outperform the traditional greedy approach, their computational cost is much higher, since the evolutionary component requires a large number (millions or billions) of function evaluations in order to produce a single tree. Aiming to reduce this computational cost, in this work we propose an encoding that allows the training and evaluation of DTs using only matrix operations. The proposed procedure is shown to be much faster than the traditional tree implementation for complete trees with depths ranging from 2 to 6, and for datasets ranging in size from 100 to 100,000 observations. In particular, the results show speedups of nearly up to 20 times, especially when the dataset is large and the desired tree is small enough to be interpretable. The proposed procedure also benefits from GPU parallelization, although it is still highly performing without it. Furthermore, we propose an evolutionary algorithm, called Coral Reef Optimization for Decision Trees (CRO-DT), that integrates this encoding with a pre-existing ensemble algorithm to evolve better univariate trees. The results obtained show that the proposed CRO-DT is competitive with traditional and modern tree algorithms, consistently producing models of good quality across 14 tested UCI Datasets. We conclude that for most relevant situations, the proposed matrix encoding provides significant speedups over the traditional implementation, and also may serve as a basis for high quality evolutionary DT algorithms. • We propose a new matrix encoding to efficiently evolve Decision Trees. • This encoding accelerates prediction through efficient numerical computations. • Great speedups are achieved for large datasets and small trees. • We introduce a tree algorithm with this encoding that uses Coral Reef Optimization. • The resulting approach searches wider and is competitive with other tree algorithms. [ABSTRACT FROM AUTHOR]

In recent years, the research community has witnessed an explosion of literature dealing with the mimicking of behavioral patterns and social phenomena observed in nature towards efficiently solving complex computational tasks. This trend has been especially dramatic in what relates to optimization problems, mainly due to the unprecedented complexity of problem instances, arising from a diverse spectrum of domains such as transportation, logistics, energy, climate, social networks, health and industry 4.0, among many others. Notwithstanding this upsurge of activity, research in this vibrant topic should be steered towards certain areas that, despite their eventual value and impact on the field of bio-inspired computation, still remain insufficiently explored to date. The main purpose of this paper is to outline the state of the art and to identify open challenges concerning the most relevant areas within bio-inspired optimization. An analysis and discussion are also carried out over the general trajectory followed in recent years by the community working in this field, thereby highlighting the need for reaching a consensus and joining forces towards achieving valuable insights into the understanding of this family of optimization techniques. [ABSTRACT FROM AUTHOR]

Abstract In medical imaging there is a special interest in relating information from different images frequently used for diagnosis or treatment. Image registration (IR) involves the transformation of different sets of image data having a shared content into a common coordinate system. The estimation of the optimal transformation is modelled either as a combinatorial or a numerical optimization problem. Since traditional IR methods are constrained by several limitations, other optimization methods have been recently proposed to overcome such shortcomings. In this contribution, we consider the use of a recently proposed and high performance bio-inspired meta-heuristic: the Coral Reef Optimization Algorithm with Substrate Layers (CRO-SL). We adapt the algorithm to the real-coding IR problem variant following both feature-based and intensity-based designs, and perform two thorough experimental studies. Such studies focus on both mono-modal and inter-modal scenarios where the images suffer different types of 3D affine transformations to validate our proposal. The new proposal is benchmarked with state-of-the-art evolutionary and non-evolutionary IR methods. The results show that CRO-SL is a very competitive approach in terms of its robustness, accuracy, and efficiency. [ABSTRACT FROM AUTHOR]

Nowadays fast-arriving information flows lay the basis of many data mining applications. Such data streams are usually affected by non-stationary events that eventually change their distribution (concept drift), causing that predictive models trained over these data become obsolete and do not adapt suitably to the new distribution. Specially in online learning scenarios, there is a pressing need for new algorithms that adapt to this change as fast as possible, while maintaining good performance scores. Recent studies have revealed that a good strategy is to construct highly diverse ensembles towards utilizing them shortly after the drift (independently from the type of drift) to obtain good performance scores. However, the existence of the so-called trade-off between stability (performance over stable data concepts) and plasticity (recovery and adaptation after drift events) implies that the construction of the ensemble model should account simultaneously for these two conflicting objectives. In this regard, this work presents a new approach to artificially generate an optimal diversity level when building prediction ensembles once shortly after a drift occurs. The approach uses a Kernel Density Estimation (KDE) method to generate synthetic data, which are subsequently labeled by means a multi-objective optimization method that allows training each model of the ensemble with a different subset of synthetic samples. Computational experiments reveal that the proposed approach can be hybridized with other traditional diversity generation approaches, yielding optimized levels of diversity that render an enhanced recovery from drifts. [ABSTRACT FROM AUTHOR]

This paper is focused on reducing the number of elements in time series with minimum information loss, with specific applications on time series segmentation. A modification of the coral reefs optimization metaheuristic (CRO) is proposed for this purpose, which is called statistical CRO (SCRO), where the main parameters of the algorithm are adjusted based on the mean and standard deviation associated with the fitness distribution. Moreover, the algorithm is combined with the Bottom-Up and Top-Down methodologies (traditional local search methods for time series segmentation), resulting in a hybrid methodology (HSCRO). We evaluate the performance of these algorithms using 16 time series from different application areas. The statistically-driven version of CRO is shown to improve the results of the standard CRO, eliminating the necessity of manually adjusting the main parameters of the algorithm and dynamically adjusting these parameters throughout the evolution. Moreover, when compared with other local search methods and metaheuristics from the state of the art, HSCRO shows robust segmentation results, consistently obtaining lower approximation errors. [ABSTRACT FROM AUTHOR]

The advent of the so-called Big Data paradigm has motivated a flurry of research aimed at enhancing machine learning models by following very diverse approaches. In this context this work focuses on the automatic construction of features in supervised learning problems, which differs from the conventional selection of features in that new characteristics with enhanced predictive power are inferred from the original dataset. In particular this manuscript proposes a new iterative feature construction approach based on a self-learning meta-heuristic algorithm (Harmony Search) and a solution encoding strategy (correspondingly, Cartesian Genetic Programming) suited to represent combinations of features by means of constant-length solution vectors. The proposed feature construction algorithm, coined as Adaptive Cartesian Harmony Search (ACHS), incorporates modifications that allow exploiting the estimated predictive importance of intermediate solutions and, ultimately, attaining better convergence rate in its iterative learning procedure. The performance of the proposed ACHS scheme is assessed and compared to that rendered by the state of the art in a toy example and three practical use cases from the literature. The excellent performance figures obtained in these problems shed light on the widespread applicability of the proposed scheme to supervised learning with legacy datasets composed by already refined characteristics. [ABSTRACT FROM AUTHOR]

Some materials, such as reinforced concrete (RC), introduce non-linear constitutive relationships into structural models. The characterization of their structural load–strain response is usually approached by using iterative strategies, such as Newton-type methods. However, these iterative strategies may not predict the non-linear structural response of RC elements (beams) due to convergence problems (non-consistent elements). In this work we propose a novel solution to overcome the convergence failures of RC structural elements subjected to shear force modeled by Compression Field Theories (CFTs). First, a Coral Reef Optimization with Substrate Layers (CRO-SL) metaheuristic is used to solve a fitting problem between numerical and experimental results on a set of real beams tested in shear, for which the system is solvable. Then, the outputs of the CRO-SL, plus the mechanical and geometric parameters of the RC beams, are used to train different Machine Learning (ML) regression methods. These regressors are then able to predict the tensile strains of the non-consistent RC elements. The characterization of such strains allows modifying the parameterization of the system of equations, overcoming thus the convergence problem of Newton-type methods in these cases. Experiments over a set of 72 real RC beams have shown the good performance of our approach. In them we have carried out a comparison among different ML regressors including Multi-Layer perceptrons, Support Vector Regression and Random Forest, plus feature reduction techniques such as PCA. [Display omitted] • Hybrid Machine Learning — Metaheuristics are proposed for improving convergence in mechanical models. • The combination of Coral Reefs Optimization algorithm with different regression models shows high performance. • Compression Field Theories are considered to predict the shear response of reinforced concrete beams. • Principal Component Analysis allows improving the shear strength experimental adjustment. [ABSTRACT FROM AUTHOR]

Highlights: [•] We propose a network clustering algorithm, based on the application of genetic operators and capable of exploiting the traffic information. [•] The algorithm can be successfully applied, rather than topology-based algorithms, to uncover relationships and detect communities whenever two nodes have a relationship though not being directly connected. [•] We compare our algorithm to other clustering algorithms, both evolutionary and non evolutionary. [•] Our algorithm outperforms all the other algorithms on all the real world datasets and outperforms all in the case of synthetic traffic matrices, excepting Newman's for the single case of a network of 75 nodes. [ABSTRACT FROM AUTHOR]

Abstract: This paper presents a hybrid evolutionary programming approach to solve the worst case tolerance design problem (WCTDP) in magnetic devices. The hybrid algorithm is formed by a basic evolutionary programming approach, mixed with a gradient-guided local search. Two different local searches procedures are tested in the paper, both specially designed to be effective in the WCTDP. Simulations on an example in the design of a magnetic circuit and comparison with several existing bio-inspired heuristics are carried out, and have shown the goodness of our algorithm. [Copyright &y& Elsevier]

Abstract: This paper provides a survey of the most important repair heuristics used in evolutionary algorithms to solve constrained optimization problems. Popular techniques are reviewed, such as some crossover operators in permutation encoding, algorithms for fixing the number of 1s in binary encoded genetic algorithms, and more specialized techniques such as Hopfield neural networks, heuristics for graphs and trees, and repair heuristics in grouping genetic algorithms. The survey also gives some indications about the design and implementation of hybrid evolutionary algorithms, and provides a revision of the most important applications in which hybrid evolutionary techniques have been used. [Copyright &y& Elsevier]

Abstract: The optimal positioning of switches in a mobile communication network is an important task, which can save costs and improve the performance of the network. In this paper we propose a model for establishing which are the best nodes of the network for allocating the available switches, and several hybrid genetic algorithms to solve the problem. The proposed model is based on the so-called capacitated p-median problem, which have been previously tackled in the literature. This problem can be split in two subproblems: the selection of the best set of switches, and a terminal assignment problem to evaluate each selection of switches. The hybrid genetic algorithms for solving the problem are formed by a conventional genetic algorithm, with a restricted search, and several local search heuristics. In this work we also develop novel heuristics for solving the terminal assignment problem in a fast and accurate way. Finally, we show that our novel approaches, hybridized with the genetic algorithm, outperform existing algorithms in the literature for the p-median problem. [Copyright &y& Elsevier]

Abstract: Handoff and cabling cost management plays a key role in the design of cellular telecommunications networks. The efficient assignment of cells to switches in this type of networks is an NP-complete problem which cannot be solved efficiently unless P=NP. This paper presents a hybrid Hopfield network-genetic algorithm approach to the cell-to-switches assignment problem, in which a Hopfield network manages the problem''s constraints, and a genetic algorithm searches for high quality solutions with the minimum possible cost in terms of handoff and cable displayed. We show, by means of computational experiments, the good performance of our approach to this problem. [Copyright &y& Elsevier]