Showing papers in "IEEE Computational Intelligence Magazine in 2018"

Recent Trends in Deep Learning Based Natural Language Processing [Review Article]

Abstract: Deep learning methods employ multiple processing layers to learn hierarchical representations of data, and have produced state-of-the-art results in many domains. Recently, a variety of model designs and methods have blossomed in the context of natural language processing (NLP). In this paper, we review significant deep learning related models and methods that have been employed for numerous NLP tasks and provide a walk-through of their evolution. We also summarize, compare and contrast the various models and put forward a detailed understanding of the past, present and future of deep learning in NLP.

Cross-Validation for Imbalanced Datasets: Avoiding Overoptimistic and Overfitting Approaches [Research Frontier]

Abstract: Although cross-validation is a standard procedure for performance evaluation, its joint application with oversampling remains an open question for researchers farther from the imbalanced data topic. A frequent experimental flaw is the application of oversampling algorithms to the entire dataset, resulting in biased models and overly-optimistic estimates.

Machine Learning for Performance Prediction in Mobile Cellular Networks

Abstract: In this paper, we discuss the application of machine learning techniques for performance prediction problems in wireless networks. These problems often involve using existing measurement data to predict network performance where direct measurements are not available. We explore the performance of existing machine learning algorithms for these problems and propose a simple taxonomy of main problem categories. As an example, we use an extensive real-world drive test data set to show that classical machine learning methods such as Gaussian process regression, exponential smoothing of time series, and random forests can yield excellent prediction results. Applying these methods to the management of wireless mobile networks has the potential to significantly reduce operational costs while simultaneously improving user experience. We also discuss key challenges for future work, especially with the focus on practical deployment of machine learning techniques for performance prediction in mobile wireless networks.

Intelligent Asset Allocation via Market Sentiment Views

Abstract: The sentiment index of market participants has been extensively used for stock market prediction in recent years. Many financial information vendors also provide it as a service. However, utilizing market sentiment under the asset allocation framework has been rarely discussed. In this article, we investigate the role of market sentiment in an asset allocation problem. We propose to compute sentiment time series from social media with the help of sentiment analysis and text mining techniques. A novel neural network design, built upon an ensemble of evolving clustering and long short-term memory, is used to formalize sentiment information into market views. These views are later integrated into modern portfolio theory through a Bayesian approach. We analyze the performance of this asset allocation model from many aspects, such as stability of portfolios, computing of sentiment time series, and profitability in our simulations. Experimental results show that our model outperforms some of the most successful forecasting techniques. Thanks to the introduction of the evolving clustering method, the estimation accuracy of market views is significantly improved.

Cluster-Based Content Distribution Integrating LTE and IEEE 802.11p with Fuzzy Logic and Q-Learning

Abstract: There is an increasing demand for distributing a large amount of content to vehicles on the road. However, the cellular network is not sufficient due to its limited bandwidth in a dense vehicle environment. In recent years, vehicular ad hoc networks (VANETs) have been attracting great interests for improving communications between vehicles using infrastructure-less wireless technologies. In this paper, we discuss integrating LTE (Long Term Evolution) with IEEE 802.11p for the content distribution in VANETs. We propose a two-level clustering approach where cluster head nodes in the first level try to reduce the MAC layer contentions for vehicle-tovehicle (V2V) communications, and cluster head nodes in the second level are responsible for providing a gateway functionality between V2V and LTE. A fuzzy logic-based algorithm is employed in the first-level clustering, and a Q-learning algorithm is used in the second-level clustering to tune the number of gateway nodes. We conduct extensive simulations to evaluate the performance of the proposed protocol under various network conditions. Simulation results show that the proposed protocol can achieve 23% throughput improvement in highdensity scenarios compared to the existing approaches.

Augmentation of Physician Assessments with Multi-Omics Enhances Predictability of Drug Response: A Case Study of Major Depressive Disorder

Abstract: This work proposes a "learning-augmented clinical assessment" workflow to sequentially augment physician assessments of patients' symptoms and their socio-demographic measures with heterogeneous biological measures to accurately predict treatment outcomes using machine learning. Across many psychiatric illnesses, ranging from major depressive disorder to schizophrenia, symptom severity assessments are subjective and do not include biological measures, making predictability in eventual treatment outcomes a challenge. Using data from the Mayo Clinic PGRN-AMPS SSRI trial as a case study, this work demonstrates a significant improvement in the prediction accuracy for antidepressant treatment outcomes in patients with major depressive disorder from 35% to 80% individualized by patient, compared to using only a physician?s assessment as the predictors. This improvement is achieved through an iterative overlay of biological measures, starting with metabolites (blood measures modulated by drug action) associated with symptom severity, and then adding in genes associated with metabolomic concentrations. Hence, therapeutic efficacy for a new patient can be assessed prior to treatment, using prediction models that take as inputs selected biological measures and physicians' assessments of depression severity. Of broader significance extending beyond psychiatry, the approach presented in this work can potentially be applied to predicting treatment outcomes for other medical conditions, such as migraine headaches or rheumatoid arthritis, for which patients are treated according to subject-reported assessments of symptom severity.

MC2ESVM: Multiclass Classification Based on Cooperative Evolution of Support Vector Machines

Abstract: Support vector machines (SVMs) are one of the most powerful learning algorithms for solving classification problems. However, in their original formulation, they only deal with binary classification. Traditional extensions of the binary SVMs for multiclass problems are based either on decomposing the problem into a number of binary classification problems, which are then independently solved, or on reformulating the objective function by solving larger optimization problems. In this paper, we propose MC2ESVM, an approach for multiclass classification based on the cooperative evolution of SVMs. Cooperative evolution allows us to decompose an M-class problem into M subproblems, which are simultaneously optimized in a cooperative fashion. We have reformulated the optimization problem such that it focuses on learning the support vectors for each class at the time that it takes into account the information from other classes. A comprehensive experimental study using common benchmark datasets is carried out to validate MC2ESVM. The experimental results, supported by statistical tests, show the effectiveness of MC2ESVM for solving multiclass classification problems, while keeping a reasonable number of support vectors.

Conflict Resolution in Mobile Networks: A Self-Coordination Framework Based on Non-Dominated Solutions and Machine Learning for Data Analytics [Application Notes]

Abstract: Self-organizing network (SON) is a well-known term used to describe an autonomous cellular network. SON functionalities aim at improving network operational tasks through the capability to configure, optimize and heal itself. However, as the deployment of independent SON functions increases, the number of dependencies between them also grows. This work proposes a tool for efficient conflict resolution based on network performance predictions. Unlike other state-of-theart solutions, the proposed self-coordination framework guarantees the right selection of network operation even if conflicting SON functions are running in parallel. This self-coordination is based on the history of network measurements, which helps to optimize conflicting objectives with low computational complexity. To do this, machine learning (ML) is used to build a predictive model, and then we solve the SON conflict by optimizing more than one objective function simultaneously. Without loss of generality, we present an analysis of how the proposed scheme provides a solution to deal with the potential conflicts between two of the most important SON functions in the context of mobility, namely mobility load balancing (MLB) and mobility robustness optimization (MRO), which require the updating of the same set of handover parameters. The proposed scheme allows fast performance evaluations when the optimization is running. This is done by shifting the complexity to the creation of a prediction model that uses historical data and that allows to anticipate the network performance. The simulation results demonstrate the ability of the proposed scheme to find a compromise among conflicting actions, and show it is possible to improve the overall system throughput.

A Fast, Adaptive, and Energy-Efficient Data Collection Protocol in Multi-Channel-Multi-Path Wireless Sensor Networks

Abstract: A wireless sensor network (WSN) consists of sensor nodes which can self-organize to relay information such as measurements to a base station. To reduce latency and increase data transmission throughput, multi-channel data collection protocols have been proposed to enable simultaneous parallel transmissions between pairs of nodes within the network. However, the existing protocols require long scheduling phase, are less dynamic to network traffic changes, and/or compromise on efficiency by relying on the back-off mechanism such as carrier sense multiple access with collision avoidance (CSMA/CA). This paper proposes a fast, adaptive, and energyefficient data collection protocol in multi-channel-multi-path WSN. The protocol consists of two major phases. The first phase is the node-channel assignment that uses the graph coloring technique to resolve the issue of node overhearing and interference. The second phase is the scheduling and packet forwarding, in which a novel three-dimensional parallel iterative matching (3DPIM) algorithm is proposed to pair up sensor nodes in different time slots so as to enable collision-free multiple simultaneous data transmissions in every time slot. Simulation results show that our proposed protocol can achieve fast and energy-efficient data collection while being adaptive to the change of network traffic in WSN.

Unsupervised Learning for Brain-Computer Interfaces Based on Event-Related Potentials: Review and Online Comparison [Research Frontier]

Abstract: One of the fundamental challenges in brain-computer interfaces (BCIs) is to tune a brain signal decoder to reliably detect a user's intention. While information about the decoder can partially be transferred between subjects or sessions, optimal decoding performance can only be reached with novel data from the current session. Thus, it is preferable to learn from unlabeled data gained from the actual usage of the BCI application instead of conducting a calibration recording prior to BCI usage. We review such unsupervised machine learning methods for BCIs based on event-related potentials of the electroencephalogram. We present results of an online study with twelve healthy participants controlling a visual speller. Online performance is reported for three completely unsupervised learning methods: (1) learning from label proportions, (2) an expectation-maximization approach and (3) MIX, which combines the strengths of the two other methods. After a short ramp-up, we observed that the MIX method not only defeats its two unsupervised competitors but even performs on par with a state-of-the-art regularized linear discriminant analysis trained on the same number of data points and with full label access. With this online study, we deliver the best possible proof in BCI that an unsupervised decoding method can in practice render a supervised method unnecessary. This is possible despite skipping the calibration, without losing much performance and with the prospect of continuous improvement over a session. Thus, our findings pave the way for a transition from supervised to unsupervised learning methods in BCIs based on eventrelated potentials.

Learning Without External Reward [Research Frontier]

Abstract: In the traditional reinforcement learning paradigm, a reward signal is applied to define the goal of the task. Usually, the reward signal is a "hand-crafted" numerical value or a pre-defined function: it tells the agent how good or bad a specific action is. However, we believe there exist situations in which the environment cannot directly provide such a reward signal to the agent. Therefore, the question is whether an agent can still learn without the external reward signal or not. To this end, this article develops a self-learning approach which enables the agent to adaptively develop an internal reward signal based on a given ultimate goal, without requiring an explicit external reward signal from the environment. In this article, we aim to convey the self-learning idea in a broad sense, which could be used in a wide range of existing reinforcement learning and adaptive dynamic programming algorithms and architectures. We describe the idealized forms of this method mathematically, and also demonstrate its effectiveness through a triple-link inverted pendulum case study.

Visualizing the Evolution of Computer Programs for Genetic Programming [Research Frontier]

Abstract: Automatically evolving computer programs to handle challenging computational problems is the main goal of genetic programming. To improve the efficiency of the evolutionary process, a large number of algorithms have been proposed in the literature.

Enhancing Selection Hyper-Heuristics via Feature Transformations

Abstract: Hyper-heuristics are a novel tool. They deal with complex optimization problems where standalone solvers exhibit varied performance. Among such a tool reside selection hyper-heuristics. By combining the strengths of each solver, this kind of hyper-heuristic offers a more robust tool. However, their effectiveness is highly dependent on the 'features' used to link them with the problem that is being solved. Aiming at enhancing selection hyper-heuristics, in this paper we propose two types of transformation: explicit and implicit. The first one directly changes the distribution of critical points within the feature domain while using a Euclidean distance to measure proximity. The second one operates indirectly by preserving the distribution of critical points but changing the distance metric through a kernel function. We focus on analyzing the effect of each kind of transformation, and of their combinations. We test our ideas in the domain of constraint satisfaction problems because of their popularity and many practical applications. In this work, we compare the performance of our proposals against those of previously published data. Furthermore, we expand on previous research by increasing the number of analyzed features. We found that, by incorporating transformations into the model of selection hyper-heuristics, overall performance can be improved, yielding more stable results. However, combining implicit and explicit transformations was not as fruitful. Additionally, we ran some confirmatory tests on the domain of knapsack problems. Again, we observed improved stability, leading to the generation of hyper-heuristics whose profit had a standard deviation between 20% and 30% smaller.

Automated Design of Machine Learning and Search Algorithms [Guest Editorial]

Abstract: 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 play an important role in solving real-world complex optimization problems in areas such as transportation, data mining, computer vision, computer security and software development, amongst others. Given the growing complexity of optimization problems, the design of effective algorithms to solve these problems has become more challenging and time consuming. The design process is itself an optimization problem. Hence, there is a demand, especially from industry and business, to automate the design process, thereby to remove the heavy reliance on human experts and to reduce the man hours involved in designing machine learning and search algorithms.

Is Evolutionary Computation Evolving Fast Enough

Abstract: Evolutionary Computation (EC) has been an active research area for over 60 years, yet its commercial/home uptake has not been as prolific as we might have expected. By way of comparison, technologies such as 3D printing, which was introduced about 35 years ago, has seen much wider uptake, to the extent that it is now available to home users and is routinely used in manufacturing. Other technologies, such as immersive reality and artificial intelligence have also seen commercial uptake and acceptance by the general public. In this paper we provide a brief history of EC, recognizing the significant contributions that have been made by its pioneers. We focus on two methodologies (Genetic Programming and Hyper-heuristics), which have been proposed as being suitable for automated software development, and question why they are not used more widely by those outside of the academic community. We suggest that different research strands need to be brought together into one framework before wider uptake is possible. We hope that this position paper will serve as a catalyst for automated software development that is used on a daily basis by both companies and home users.

Fast Unsupervised Edge Detection Using Genetic Programming [Application Notes]

Abstract: Edge detection has been a fundamental and important task in computer vision for many years, but it is still a challenging problem in real-time applications, especially for unsupervised edge detection, where ground truth is not available. Typical fast edge detection approaches, such as the single threshold method, are expensive to achieve in unsupervised edge detection. This study proposes a Genetic Programming (GP) based algorithm to quickly and automatically extract binary edges in an unsupervised manner. We investigate how GP can effectively evolve an edge detector from a single image without ground truth, and whether the evolved edge detector can be directly applied to other unseen/test images. The proposed method is examined and compared with a recent GP method and the Canny method on the Berkeley segmentation dataset. The results show that the proposed GP method has the ability to effectively evolve edge detectors by using only a single image as the whole training set, and significantly outperforms the two methods it is compared to. Furthermore, the binary edges detected by the evolved edge detectors have a good balance between recall and precision.

Computational Intelligence Techniques for Mobile Network Optimization [Guest Editorial]

Abstract: Modern society has become increasingly reliant on mobile networks for their communication needs. Such networks are characterized by their dynamic, heterogeneous, complex, and data intensive nature, which makes them more amenable to automated mobile network optimization performed using “computational intelligence’’ (CI) techniques rather than traditional optimization approaches. CI techniques—which subsume multidisciplinary techniques from machine learning (ML), optimization theory, game theory, control theory, and meta-heuristics— have a rich history in terms of being deployed in networking. CI techniques are highly suited to the mobile networking architectures and the dynamic environments they characterize. Looking ahead, it looks likely that CI will play a leading role in upcoming 5th generation (5G) wireless mobile networks for developing optimized solutions for vexing problems—such as traffic scheduling and routing, capacity, coverage, and power optimization—in the face of stringent requirements and highly dynamic conditions. The importance of our proposed theme of mobile network optimization (MNO) motivated us to propose this special issue in the IEEE Computational Intelligence Magazine (CIM)—the premier IEEE magazine for professionals interested in CI techniques and their applications.

Optimal Weighted Extreme Learning Machine for Imbalanced Learning with Differential Evolution [Research Frontier]

Abstract: In this paper, we present a formal model for the optimal weighted extreme learning machine (ELM) on imbalanced learning. Our model regards the optimal weighted ELM as an optimization problem to find the best weight matrix. We propose an approximate search algorithm, named weighted ELM with differential evolution (DE), that is a competitive stochastic search technique, to solve the optimization problem of the proposed formal imbalanced learning model. We perform experiments on standard imbalanced classification datasets which consist of 39 binary datasets and 3 multiclass datasets. The results show a significant performance improvement over standard ELM with an average Gmean improvement of 10.15% on binary datasets and 1.48% on multiclass datasets, which are also better than other state-of-the-art methods. We also demonstrate that our proposed algorithm can achieve high accuracy in representation learning by performing experiments on MNIST, CIFAR-10, and YouTube-8M, with feature representation from convolutional neural networks.

Market Model Benchmark Suite for Machine Learning Techniques

Abstract: Recent developments in deep-reinforcement learning have yielded promising results in artificial games and test domains. To explore opportunities and evaluate the performance of these machine learning techniques, various benchmark suites are available, such as the Arcade Learning Environment, rllab, OpenAI Gym, and the StarCraft II Learning Environment. This set of benchmark suites is extended with the open business simulation model described here, which helps to promote the use of machine learning techniques as valueadding tools in the context of strategic decision making and economic model calibration and harmonization. The benchmark suite extends the current state-of-the-art problems for deep-reinforcement learning by offering an infinite state and action space for multiple players in a non-zero-sum game environment of imperfect information. It provides a model that can be characterized as both a credit assignment problem and an optimization problem. Experiments with this suite?s deep-reinforcement learning algorithms, which yield remarkable results for various artificial games, highlight that stylized market behavior can be replicated, but the infinite action space, simultaneous decision making, and imperfect information pose a computational challenge. With the directions provided, the benchmark suite can be used to explore new solutions in machine learning for strategic decision making and model calibration.

Computational Intelligence Techniques in Bioinformatics and Bioengineering [Guest Editorial]

Abstract: The computational intelligence community in bioinformatics and bioengineering is fragmented and large The aim of the special issue is to capture some of the ongoing interdisciplinary research that draws upon joint expertise in the domains of optimization and learning via computational intelligence techniques and bioinformatics and bioengineering After a rigorous review process, two papers were selected for publication in the special issue The papers are briefly summarized

An Accurate Lattice Model for Pricing Catastrophe Equity Put Under the Jump-Diffusion Process

Abstract: A catastrophe equity put (CatEPut) is constructed to recapitalize an insurance company that suffers huge compensation payouts due to catastrophic events (CEs). The company can exercise its CatEPut to sell its stock to the counterparty at a predetermined price when its accumulated loss due to CEs exceeds a predetermined threshold and its own stock price falls below the strike price

Identifying DNA Methylation Modules Associated with a Cancer by Probabilistic Evolutionary Learning

Abstract: DNA methylation leads to inhibition of downstream gene expression. Recently, considerable studies have been made to determine the effects of DNA methylation on complex disease. However, further studies are necessary to find the multiple interactions of many DNA methylation sites and their association with cancer. Here, to assess DNA methylation modules potentially relevant to disease, we use an Estimation of Distribution Algorithm (EDA) to identify high-order interaction of DNA methylated sites (or modules) that are potentially relevant to disease. The method builds a probabilistic dependency model to produce a solution that is a set of discriminative methylation sites. The algorithm is applied to array- and sequencing-based high-throughput DNA methylation profiling datasets. The experimental results show that it is able to identify DNA methylation modules for cancer.

IEEE CIM Survey Results [Editor's Remarks]

Abstract: Presents the results of a CIM Web-based survey which was sent out to about 4,000 CIS society members asking for their opinions and suggestions to enhance and improve IEEE Computational Intelligence Magazine.