Subscribe: IEEE Intelligent Systems
Added By: Feedage Forager Feedage Grade B rated
Language: English
analysis  approach  article  authors  based  challenges  data  feature  method  model  proposed  results  system  systems  users 
Rate this Feed
Rate this feedRate this feedRate this feedRate this feedRate this feed
Rate this feed 1 starRate this feed 2 starRate this feed 3 starRate this feed 4 starRate this feed 5 star

Comments (0)

Feed Details and Statistics Feed Statistics
Preview: IEEE Intelligent Systems

IEEE Intelligent Systems

IEEE Intelligent Systems is a bimonthly publication of the IEEE Computer Society that provides peer-reviewed, cutting-edge articles on the theory and applications of systems that perceive, reason, learn, and act intelligently. The editorial staff collabor


Fully Autonomous Driving: Where Technology and Ethics Meet

10/19/2017 2:40 pm PST

The prospect of automatized car traffic confronts ethics, law, and politics with novel and far-reaching questions. It cannot be excluded that automatized cars meet with critical situations in which losses of life and limb are inevitable and in which the necessity arises to negotiate between two or more evils. Despite all security measures, a residual risk is unavoidable, which raises questions: How safe is safe enough? How safe is too safe? This contribution argues that it is important to clearly separate between the tasks of technology and ethics as well as between the responsibilities of different stakeholders.

Robust and Fast Segmentation Based on Fuzzy Clustering Combined with Unsupervised Histogram Analysis

10/19/2017 2:39 pm PST

Many real-time engineering applications have used histogram thresholding methods that failed to segment images whose histogram had only one peak. A fuzzy c-means cluster algorithm (FCM), in contrast, can segment this type of image but at the cost of time. To improve unsupervised segmentation, the authors developed a new method for fast and efficient segmentation based on automatic histogram analysis of acquired images and a combined FCM and intensity transformation (HIST_FCM_IT) approach. The first part of the algorithm uses parabolic approximation for peak evaluation, and the second modifies image intensity to allow the partition matrix to be rapidly constant.

Design and Prototyping a Smart Deep Brain Stimulator: An Autonomous Neuro-Sensing and Stimulating Electrode System

01/11/2018 10:11 am PST

This article presents the design and prototyping of an innovative smart deep brain stimulator (SDBS) that consists of brain-implantable smart electrodes and a wireless-connected external controller. SDBS electrodes operate as completely autonomous electronic implants that are capable of sensing and recording neural activities in real time, performing local processing, and generating arbitrary waveforms for neuro-stimulation. A bidirectional, secure, fully passive wireless communication backbone was designed and integrated into this smart electrode to maintain contact between the electrodes and the controller. The standard EPC-Global protocol has been modified and adopted as the communication protocol in this design. The proposed SDBS was demonstrated and tested through a hardware prototypes.

SOM-Optimized Neurofuzzy Classifiers for Measuring Expatriation Willingness

10/19/2017 2:40 pm PST

The authors developed a self-organizing map-based optimization (SOMO) neurofuzzy classifier to deal with a practical expatriation willingness (EW) problem, which is associated with employees’ willingness to accept expatriate assignments. The proposed model also delivers more information about rule coverage and generates user-friendly outcomes. The authors adopt the SOMO algorithm to optimize the weights of neurons in the proposed hybrid neurofuzzy classifier. They evaluated the model’s feasibility using the databases from previous studies exploring employees’ EW. The results show that the proposed neurofuzzy classifier yields 11 determination rules whose accuracy rates are greater than 80 percent. The contribution to the body of knowledge lies in the significant improvement in accuracy rates and coverage for predicting EW rules, optimization for all the network parameters using a novel algorithm, and more friendly outputs for practical use.

A Set Space Model for Feature Calculus

10/19/2017 2:39 pm PST

Processing natural language at the sentence level suffers from a sparse-feature problem caused by the limited number of words in a sentence. In this article, a Set Space Model (SSM) is proposed to utilize sentence information, the main idea being that, depending on structural characteristics or functional principles of linguistics, features in a sentence can be grouped into different sets. Feature calculus can then operate on the grouped features and capture structural information using external knowledge. The authors implement this method in a traditional information extraction task, with results showing significant and constant improvement in general information extraction.

The Inductive Constraint Programming Loop

10/19/2017 2:39 pm PST

Constraint programming is used for a variety of real-world optimization problems, such as planning, scheduling, and resource allocation problems, all while we continuously gather vast amounts of data about these problems. Current constraint programming software doesn’t exploit such data to update schedules, resources, and plans. The authors propose a new framework that they call the inductive constraint programming loop. In this approach, data is gathered and analyzed systematically to dynamically revise and adapt constraints and optimization criteria. Inductive constraint programming aims to bridge the gap between the areas of data mining and machine learning on one hand and constraint programming on the other.

Collective Hyping Detection System for Identifying Online Spam Activities

12/11/2017 4:11 pm PST

Although existing antispam strategies detect traditional spam activities effectively, evolving spam schemes can successfully cheat conventional testing by buying the comments that are written by genuine users and sold by specific web markets. Such spam activities turn into a kind of advertising campaign among business owners to maintain their rank in top positions. This article proposes a new collaborative marketing hyping detection solution that aims to identify spam comments generated by the Spam Reviewer Cloud and detect products that adopt an evolving spam strategy for promotion. The authors propose an unsupervised learning model that combines heterogeneous product review networks in an attempt to discover collective hyping activities. Their experiments validate the existence of the collaborative marketing hyping activities on a real-life ecommerce platform and demonstrate that their model can effectively and accurately identify these advanced spam activities.

On Using the Intelligent Edge for IoT Analytics

10/19/2017 2:40 pm PST

This article presents a flexible architecture for Internet of Things (IoT) data analytics using the concept of fog computing. The authors identify different actors and their roles in order to design adaptive IoT data analytics solutions. The presented approach can be used to effectively design robust IoT applications that require a tradeoff between cloud- and edge-based computing depending on dynamic application requirements. The potential use cases of this technology can be found in scenarios such as smart cities, security surveillance, and smart manufacturing, where the quality of user experience is important.

Challenges of Sentiment Analysis for Dynamic Events

10/19/2017 2:39 pm PST

Efforts to assess people’s sentiments on Twitter have suggested that Twitter could be a valuable resource for studying political sentiment and that it reflects the offline political landscape. Many opinion mining systems and tools provide users with people’s attitudes toward products, people, or topics and their attributes/aspects. However, although it may appear simple, using sentiment analysis to predict election results is difficult, since it is empirically challenging to train a successful model to conduct sentiment analysis on tweet streams for a dynamic event such as an election. This article highlights some of the challenges related to sentiment analysis encountered during monitoring of the presidential election using Kno.e.sis’s Twitris system.

Challenges in User Modeling and Personalization

10/19/2017 2:40 pm PST

Personalization has a long history, dating back to the “master-apprentice” approach of individual tutoring that sought to pass on knowledge and skills from one generation to the next. Through user modeling and adaptation, we try to capture the tutor’s human intelligence and turn it into artificial intelligence. Over the last decades, this research has evolved from an expert-driven approach toward a data-driven approach. This evolution comes with an interesting challenge: How can we continue to understand what an automated tutor is doing when the process of collecting and interpreting data about users is fully automated and the adaptation and recommendation decisions are “deduced” from individual users’ behavior as well as the behavior of all users combined? This article discusses the challenges of scrutability, repeatability, and meta-adaptation (aka adaptation of the adaptation), important research issues for the coming years.

Cognitive Computing

08/21/2017 2:49 pm PST

The guest editors of this special issue on cognitive computing discuss the field in general and the four articles they selected to represent it in particular.

Interactive Task Learning

08/21/2017 2:49 pm PST

This article presents a new research area called interactive task learning (ITL), in which an agent actively tries to learn not just how to perform a task better but the actual definition of a task through natural interaction with a human instructor while attempting to perform the task. The authors provide an analysis of desiderata for ITL systems, a review of related work, and a discussion of possible application areas for ITL systems.

Interactive Cognitive Systems and Social Intelligence

08/21/2017 2:49 pm PST

Research on cognitive systems adopts the aims and assumptions of classical AI research, emphasizing the construction of intelligent agents that exhibit complex behavior. This article reviews the cognitive systems paradigm and two widely adopted hypotheses—physical symbol systems and heuristic search—that underpin it. The author introduces a third claim—the social cognition hypothesis—that states intelligence requires the ability to represent and reason about others’ mental states. The article also examines a number of computational artifacts, both historical and recent, that focus on interaction and exhibit this capacity. Examples include dialogue systems, synthetic experts, believable agents, intelligent tutors, interactive robots, and instructable game players. In closing, the author identifies issues in social cognition that deserve greater attention and poses challenges that can drive future research on interactive cognitive systems.

Computers Play Chess, Computers Play Go…Humans Play Dungeons & Dragons

11/22/2017 10:05 am PST

With the AlphaGo computer program’s recent win over one of the world’s expert Go players, AI researchers need to explore new challenges in the game-playing arena. While there are a number of games to explore, the authors pose a true challenge for the next decade: attacking human-oriented games such as Dungeons & Dragons.

Meta-Algorithms in Cognitive Computing

08/21/2017 2:49 pm PST

A brief overview of algorithms that improve other algorithms and their role in cognitive computing.

Using Process Mining to Model Multi-UAV Missions through the Experience

08/21/2017 2:49 pm PST

The interest in civilian missions with multiple unmanned aerial vehicles (UAVs) has increased significantly in recent years, but these missions pose multiple challenges related to operator workload and situational awareness. Human-machine interfaces must consider these challenges and control the amount of information flowing to operators. The authors propose a procedure for automatically modeling multi-UAV missions that uses process mining to discover Petri nets through event logs. Specifically, it applies several discovery algorithms, generates and evaluates models, and determines the best among the four.

Data Mining-Based Decomposition for Solving the MAXSAT Problem: Toward a New Approach

08/21/2017 2:49 pm PST

This article explores advances in the data mining arena to solve the fundamental MAXSAT problem. In the proposed approach, the MAXSAT instance is first decomposed and clustered by using data mining decomposition techniques, then every cluster resulting from the decomposition is separately solved to construct a partial solution. All partial solutions are merged into a global one, while managing possible conflicting variables due to separate resolutions. The proposed approach has been numerically evaluated on DIMACS instances and some hard Uniform-Random-3-SAT instances, and compared to state-of-the-art decomposition based algorithms. The results show that the proposed approach considerably improves the success rate, with a competitive computation time that’s very close to that of the compared solutions.

Consumption Behavior Analytics-Aided Energy Forecasting and Dispatch

08/21/2017 2:50 pm PST

For decades, electricity customers have been treated as mere recipients of electricity in vertically integrated power systems. However, as customers have widely adopted distributed energy resources and other forms of customer participation in active dispatch (such as demand response) have taken shape, the value of mining knowledge from customer behavior patterns and using it for power system operation is increasing. Further, the variability of renewable energy resources has been considered a liability to the grid. However, electricity consumption has shown the same level of variability and uncertainty, and this is sometimes overlooked. This article investigates data analytics and forecasting methods to identify correlations between electricity consumption behavior and distributed photovoltaic (PV) output. The forecasting results feed into a predictive energy management system that optimizes energy consumption in the near future to balance customer demand and power system needs.

Toward a Machine Intelligence Layer for Diverse Industrial IoT Use Cases

08/21/2017 2:49 pm PST

The Internet of Things (IoT) has moved beyond the hype, with promising applications materializing and industries transforming through digitalization as well as servitization—that is, delivering a service as an integral part of a product. Since the application spread in today’s IoT is wide and is typically structured in market-oriented groups, a system designer needs IoT system design patterns to assist in designing for scalable and replicable solutions. The work presented here provides a generic blueprint for designers to jumpstart the design process of an unknown use case.

Sentiment Analysis in TripAdvisor

10/12/2017 3:52 pm PST

Web platforms such as TripAdvisor allow tourists to describe their experiences with hotels, restaurants, and other tourist attractions. This article proposes TripAdvisor as a source of data for sentiment analysis tasks. The authors develop an analysis for studying the matching between users’ sentiments and automatic sentiment-detection algorithms. They also discuss some of the challenges regarding sentiment analysis and TripAdvisor.

Explaining Explanation, Part 2: Empirical Foundations

08/21/2017 2:49 pm PST

This article surveys empirical research that reveals the variety and diversity of the forms and purposes of causal reasoning, and reveals the myths that have driven philosophical analysis, psychological research, and computational approaches. The intent of the essay is to broaden the horizons for the development of intelligent systems that serve explanatory functions.

Building DARPA’s Brain

08/21/2017 2:50 pm PST

DARPA is helping to lead a remarkable transition from traditional human-centric creativity to a new kind of innovation that emerges from symbiotic collaboration among humans and machines. This evolution, the result of fast-paced changes in artificial intelligence and machine learning, is already offering new and better approaches to designing devices and manipulating the world around us, but is also raising novel questions about human-machine relationships including questions about validation and trust. A few DARPA research efforts are described that exemplify some of the benefits and quandaries that are emerging in this new era of human-machine accelerated discovery.

IEEE Computer Society 2017 Call for Major Award Nominations

08/21/2017 2:49 pm PST

Presents a call for nominations for select CS society awards.

Computational Advertising: A Paradigm Shift for Advertising and Marketing?

06/12/2017 2:34 pm PST

The umbrella term "computational advertising" encompasses a spectrum of computational systems, technologies, and methods of advertising and promotional behaviors and decision-making activities. The guest editors of this special issue on computational advertising discuss the field in general and the highlighted articles in particular.

Utilizing Verbal Intent in Semantic Contextual Advertising

06/12/2017 2:45 pm PST

Existing semantic approaches to contextual advertising effectively match relevant ads to webpages in terms of topical intent. The authors observe, however, that there's still much room for improvement. In this article, they seek to utilize the verbal intent, which complements topical intent, in semantic contextual advertising. Verbal intent describes what a user wants to do, that is, the action perspective, with topical intent. The authors propose a methodology that effectively identifies verbal intent in webpages and ads, and then incorporates the verbal intent into an ad-ranking framework. The results of performance evaluation on real-world datasets clearly show the efficacy of (verb, topic) associations. Compared with state-of-the-art techniques, the proposed methodology exhibits a significant improvement of precision at a high level of recall in ad ranking, as well as a precision improvement of 35 percent on average in verb identification.

AdScope: Search Campaign Scoping Using Relevance Feedback

06/12/2017 2:45 pm PST

In a search ad campaign, the host search network provides information on expenses and revenue, user clicks, conversions, and search queries issued pre-click. An experienced advertiser goes through these queries and identifies the relevant ones to sharpen and the irrelevant ones to shrink to improve the campaign's reach and scope. With the right scope, the budget can be spent to target relevant users. AdScope ranks user queries with respect to relevance and recommends to advertisers the topmost queries for inclusion in a campaign and the bottommost queries for exclusion. It does this by combining the feedback collected from both users and advertisers to improve the ranking. The authors measured AdScope's performance with relevance classification and found that it achieved the highest classification accuracy (89.3) percent for queries that contain at least two terms.

A Three-Phase Approach for Exploiting Opinion Mining in Computational Advertising

08/30/2017 8:59 am PST

In the past decade, advertising has rapidly become a major profit source for web-based businesses. One of the main challenges is the creation of effective and engaging ads that attract potential customers to product websites and eventual purchase. A promising solution is the combination of user preferences from the content provided on social media and the extraction of significant aspects discussed on review websites. The author proposes a three-phase model for content analysis on product review sites that considers in tandem the aspects discussed by users on review websites, the opinion polarities associated with such aspects, and user profiles on social networks, aiming to detect the most "interesting" aspects and use them to generate attractive messages. The effectiveness of the proposed approach has been validated in a real-world scenario by working with two Twitter seeds about domain-specific magazines and the data of their respective followers. Results demonstrate how the messages created by analyzing content (that is, product aspects and associated polarities) provided by a specific community improves the overall attractiveness of the generated advertisements.

Exploring Interpersonal Influence by Tracking User Dynamic Interactions

07/06/2017 11:56 am PST

As one of the most popular social media platforms today, Twitter provides people with an effective way to communicate and interact with each other. Through these interactions, influence among users gradually emerges and changes people's opinions. Although previous work has studied interpersonal influence as the probability of activating others during information diffusion, they ignore an important fact that information diffusion is the result of influence, while dynamic interactions among users produce influence. In this article, the authors propose a novel temporal influence model to learn users' opinion behaviors regarding a specific topic by exploring how influence emerges during communications. The experiments show that their model performs better than other influence models with different influence assumptions when predicting users' future opinions, especially for the users with high opinion diversity.

Learning Geographical and Mobility Factors for Mobile Application Recommendation

06/12/2017 2:45 pm PST

With myriad features and functionalities, mobile app users have the option to run different types of apps when they move to different locations. For a specific place, the decision process involved in choosing a mobile app can be complex and influenced by various factors, such as app popularity, user preferences, geographical influences, and user mobility behaviors. Although several researchers have studied recommendation in mobile apps, they omitted an integrated analysis of the joint effect of multiple factors from a geographical perspective. This article proposes a novel location-based probabilistic factor analysis mechanism that considers multiple factors to help people visiting a new location get an appropriate mobile app recommendation. In particular, the authors model mobile app usage from a geographical perspective. Experimental results on real mobile usage data show that the proposed recommendation method outperforms baseline algorithms by 30 percent.

Game-Theoretic Considerations for Optimizing Taxi System Efficiency

06/12/2017 2:33 pm PST

Taxi service is an indispensable part of public transport in modern cities. To support its unique features, a taxi system adopts a decentralized operation mode in which thousands of taxis freely decide their working schedules and routes. Taxis compete with each other for individual profits regardless of system-level efficiency, making the taxi system inefficient and hard to optimize. Most research into the management and economics of taxi markets has focused on modeling from a macro level the effects of and relationships between various market factors. Less has been done regarding a more important component--drivers' strategic behavior under the decentralized operation mode. The authors propose looking at the problem from a game-theoretic perspective. Combining game-theoretic solution concepts with existing models of taxi markets, they model taxi drivers' strategy-making process as a game and transform the problem of optimizing taxi system efficiency into finding a market policy that leads to the desired equilibrium.

5G-Enabled Cooperative Intelligent Vehicular (5GenCIV) Framework: When Benz Meets Marconi

06/12/2017 2:45 pm PST

As one of the most popular social media platforms today, Twitter provides people with an effective way to communicate and interact with each other. Through these interactions, influence among users gradually emerges and changes people's opinions. Although previous work has studied interpersonal influence as the probability of activating others during information diffusion, they ignore an important fact that information diffusion is the result of influence, while dynamic interactions among users produce influence. In this article, the authors propose a novel temporal influence model to learn users' opinion behaviors regarding a specific topic by exploring how influence emerges during communications. The experiments show that their model performs better than other influence models with different influence assumptions when predicting users' future opinions, especially for the users with high opinion diversity.

GAIA: A CAD-Like Environment for Designing Game-Playing Agents

06/12/2017 2:45 pm PST

CAD environments enable designers to construct, evaluate, and revise models of engineering systems. GAIA is a CAD-like environment for designing game-playing agents. Unlike engineering systems, intelligent agents may learn from experience. Thus, in GAIA, the human designer and the intelligent agent cooperate to redesign the agent. In this article, the authors describe three elements of this vision: the interactive environment GAIA, an agent modeling language called TMKL2, and a GAIA module called REM that performs meta-reasoning for self-adaptation in game-playing agents. They illustrate these concepts for designing software agents that play variants of Freeciv, a turn-based strategy game.

Explaining Explanation, Part 1: Theoretical Foundations

06/12/2017 2:45 pm PST

This is the first in a series of essays that addresses the manifest programmatic interest in developing intelligent systems that help people make good decisions in messy, complex, and uncertain circumstances by exploring several questions: What is an explanation? How do people explain things? How might intelligent systems explain their workings? How might intelligent systems help humans be better understanders as well as better explainers? This article addresses the theoretical foundations.

Manufacturing Analytics and Industrial Internet of Things

12/13/2017 1:02 pm PST

Over the last two decades, manufacturing across the globe has evolved to be more intel-ligent and data driven. In the age of industrial Internet of Things, a smart production unit can be perceived as a large connected industrial system of materials, parts, machines, tools, inventory, and logistics that can relay data and communicate with each other. While, traditionally, the focus has been on machine health and predictive maintenance, the manufacturing industry has also started focusing on analyzing data from the entire production line. These applications bring a new set of analytics challenges. Unlike tradi-tional data mining analysis, which consists of lean datasets (that is, datasets with few fea-tures), manufacturing has fat datasets. In addition, previous approaches to manufacturing analytics restricted themselves to small time periods of data. The latest advances in big data analytics allows researchers to do a deep dive into years of data. Bosch collects and utilizes all available information about its products to increase its understanding of complex linear and nonlinear relationships between parts, machines, and assembly lines. This helps in use cases such as the discovery of the root cause of internal defects. This article presents a case study and provides detail about challenges and approaches in data extraction, modeling, and visualization.

Aspect-Based Extraction and Analysis of Affective Knowledge from Social Media Streams

06/12/2017 2:45 pm PST

Extracting and analyzing affective knowledge from social media in a structured manner is a challenging task. Decision makers require insights into the public perception of a company's products and services, as a strategic feedback channel to guide communication campaigns, and as an early warning system to quickly react in the case of unforeseen events. The approach presented in this article goes beyond bipolar metrics of sentiment. It combines factual and affective knowledge extracted from rich public knowledge bases to analyze emotions expressed toward specific entities (targets) in social media. The authors obtain common and common-sense domain knowledge from DBpedia and ConceptNet to identify potential sentiment targets. They employ affective knowledge about emotional categories available from SenticNet to assess how those targets and their aspects (such as specific product features) are perceived in social media. An evaluation shows the usefulness and correctness of the extracted domain knowledge, which is used in a proof-of-concept data analytics application to investigate the perception of car brands on social media in the period between September and November 2015.

In Memoriam: Adele Howe

03/30/2017 2:36 pm PST

Recounts the career and contributions of Adele Howe.

2016 Reviewer Thanks

03/30/2017 2:34 pm PST

The editorial team (staff and volunteer) thanks the multiple people who perform the valuable service of peer review on the magazine's article and department content.

Big Data

03/30/2017 2:36 pm PST

Big data has been an enabler for innovation, reconstruction, and advancement of most sectors of our society, and it's receiving continuous and growing attention from researchers and practitioners in academia, industry, and government. There are, however, still lots of challenges spanning from theoretical foundations, systems, and technology to data policy and standards. This special issue focuses on how big data cuts across systems and applications arenas, and the guest editors' introduction describes the five articles they selected out of 30 submitted to cover a wide spectrum of interesting topics, including feature selection for big data analytics, astronomical image analysis, large-scale network prediction, online URL filtering, and massive transaction clustering.

Challenges of Feature Selection for Big Data Analytics

03/30/2017 2:36 pm PST

We're surrounded by huge amounts of large-scale high-dimensional data, but learning tasks require reduced data dimensionality. Feature selection has shown its effectiveness in many applications by building simpler and more comprehensive models, improving learning performance, and preparing clean, understandable data. Some unique characteristics of big data such as data velocity and data variety have presented challenges to the feature selection problem. In this article, the authors envision these challenges for big data analytics. To facilitate and promote feature selection research, they present an open source feature selection repository (scikit-feature) of popular algorithms.

Big Universe, Big Data: Machine Learning and Image Analysis for Astronomy

03/30/2017 2:35 pm PST

Astrophysics and cosmology are rich with data. The advent of wide-area digital cameras on large aperture telescopes has led to ever more ambitious surveys of the sky. Data volumes of entire surveys a decade ago can now be acquired in a single night, and real-time analysis is often desired. Thus, modern astronomy requires big data know-how, in particular, highly efficient machine learning and image analysis algorithms. But scalability isn't the only challenge: astronomy applications touch several current machine learning research questions, such as learning from biased data and dealing with label and measurement noise. The authors argue that this makes astronomy a great domain for computer science research, as it pushes the boundaries of data analysis. They focus here on exemplary results, discuss main challenges, and highlight some recent methodological advancements in machine learning and image analysis triggered by astronomical applications.

Structured Regression on Multiscale Networks

03/30/2017 2:36 pm PST

Structure-based regression algorithms generally suffer substantive speed losses and have exacting memory requirements compared to their structureless counterparts. Gaussian conditional random field (GCRF) models are one of the most time- and memory-efficient approaches to structured regression. The authors' previous speedups for the GCRF method allow for exact solutions on networks of up to 100,000 nodes and 10 million links. Using multiscale networks, the exact solution for networks of millions of nodes and trillions of links can be solved in a similar amount of time. They walk through the intuitiveness of using multiple scales of networks on a real-life health informatics application. The time and memory demands from using this approach are logarithmic compared to naive implementations.

Online URL Classification for Large-Scale Streaming Environments

03/30/2017 2:36 pm PST

Large-scale streaming URLs are the norm in many commercial software products that aim to filter URLs based on their sensitivity or risk level. In such problem scenarios, filtering is typically done by classifying a URL using either its webpage content or certain additional contextual information. However, such approaches are slow and computationally expensive, as they require gathering and processing webpage content or other contextual information for each URL. In this work, the authors propose a method for classifying URLs in large-scale streaming environments that doesn't suffer from these drawbacks. The proposed method is based on online ensemble learning, which results in lightweight prediction models that are well-suited for classification of streaming datasets. The authors illustrate the effectiveness of the proposed approach using large-scale datasets from a live, production environment and show that the proposed method results in an increase of 5 to 8 percent in terms of precision and 3 to 5.5 percent in terms of recall.

Local PurTree Spectral Clustering for Massive Customer Transaction Data

03/30/2017 2:36 pm PST

The clustering of customer transaction data is very important to retail and e-commerce companies. The authors propose a local PurTree spectral clustering algorithm for massive customer transaction data that uses a purchase tree to represent customer transaction data and a PurTree distance to compute the distance between two trees. The new method learns a data similarity matrix from the local distances and the level weights in the PurTree distance simultaneously. An iterative optimization algorithm is proposed to optimize the proposed model. The authors conducted experiments to compare their method with four commonly used clustering method for transaction data on six real-life datasets. The experimental results show that the new method outperformed other clustering algorithms.

Collaborative Recommendation with Multiclass Preference Context

03/30/2017 2:35 pm PST

Factorization- and neighborhood-based methods have been recognized as state-of-the-art approaches for collaborative recommendation tasks. In this article, the authors take user ratings as categorical multiclass preferences and propose a novel method called matrix factorization with multiclass preference context (MF-MPC), which integrates an enhanced neighborhood based on the assumption that users with similar past multiclass preferences (instead of one-class preferences in SVD++) will have similar tastes in the future. The main merit of MF-MPC is its ability to make use of the multiclass preference context in the factorization framework in a fine-grained manner and thus inherit the advantages of those two methods. Experimental results on three real-world datasets show that their solution can perform significantly better than factorization-based methods, neighborhood-based methods, and integrated methods with a one-class preference context.

A Hybrid Approach for the Sudoku Problem: Using Constraint Programming in Iterated Local Search

03/30/2017 2:35 pm PST

Sudoku is not only a popular puzzle but also an interesting and challenging constraint satisfaction problem. Therefore, automatic solving methods have been the subject of several publications in the past two decades. Although current methods provide good solutions for small-sized puzzles, larger instances remain challenging. This article introduces a new local search technique based on the min-conflicts heuristic for Sudoku. Furthermore, the authors propose an innovative hybrid search technique that exploits constraint programming as a perturbation technique within the iterated local search framework. They experimentally evaluate their methods on challenging benchmarks for Sudoku and report improvements over state-of-the-art solutions. To show the generalizability of the proposed approach, they also applied their method on another challenging scheduling problem. The results show that the proposed method is also robust in another problem domain.

Emerging White-Collar Robotics: The Case of Watson Analytics

03/30/2017 2:36 pm PST

This article discusses easy and smart analytics through the example of Watson Analytics. The author reviews Watson Analytics capabilities, summarizes its strengths and limitations, and analyzes data in two case studies.

IoT Quality Control for Data and Application Needs

03/30/2017 2:35 pm PST

The amount of Internet of Things (IoT) data is growing rapidly. Although there is a growing understanding of the quality of such data at the device and network level, important challenges in interpreting and evaluating the quality at informational and application levels remain to be explored. This article discusses some of these challenges and solutions of IoT systems at the different OSI layers to understand the factors affecting the quality of the overall system. With the help of two IoT-enabled digital health applications, the authors investigate the role of semantics in measuring the data quality of the system, as well as integrating multimodal data for clinical decision support. They also discuss the extension of IoT to the Internet of Everything by including human-in-the-loop to enhance the system accuracy. This paradigm shift through the confluence of sensors and data analytics can lead to accelerated innovation in applications by overcoming the limitations of the current systems, leading to unprecedented opportunities in healthcare.

Deep Learning-Based Document Modeling for Personality Detection from Text

03/30/2017 2:35 pm PST

This article presents a deep learning based method for determining the author's personality type from text: given a text, the presence or absence of the Big Five traits is detected in the author's psychological profile. For each of the five traits, the authors train a separate binary classifier, with identical architecture, based on a novel document modeling technique. Namely, the classifier is implemented as a specially designed deep convolutional neural network, with injection of the document-level Mairesse features, extracted directly from the text, into an inner layer. The first layers of the network treat each sentence of the text separately; then the sentences are aggregated into the document vector. Filtering out emotionally neutral input sentences improved the performance. This method outperformed the state of the art for all five traits, and the implementation is freely available for research purposes.

Graph Structure Learning from Unlabeled Data for Early Outbreak Detection

03/30/2017 2:35 pm PST

Processes such as disease propagation and information diffusion often spread over some latent network structure that must be learned from observation. Given a set of unlabeled training examples representing occurrences of an event type of interest (such as a disease outbreak), the authors aim to learn a graph structure that can be used to accurately detect future events of that type. They propose a novel framework for learning graph structure from unlabeled data by comparing the most anomalous subsets detected with and without the graph constraints. Their framework uses the mean normalized log-likelihood ratio score to measure the quality of a graph structure, and it efficiently searches for the highest-scoring graph structure. Using simulated disease outbreaks injected into real-world Emergency Department data from Allegheny County, the authors show that their method learns a structure similar to the true underlying graph, but enables faster and more accurate detection.

The Golden Age of AI

02/11/2017 2:32 pm PST

Incoming editor-in-chief V.S. Subrahmanian describes the state of AI and his plans for the magazine.

AI and Economics [Guest editors' introduction]

02/16/2017 3:04 pm PST

The fields of economic theory and artificial intelligence (AI) share many common interests and notions such as utility, probability, and reasoning about other actors in the environment. This common ground has given rise to a wide body of literature at the interface of economic theory and AI whose scope includes algorithmic aspects of cooperative and noncooperative game theory, mechanism design, and computational social choice, just to name a few. This special issue reports the state of the art in theory, algorithms, and applications in this broad area.

Agent Failures in All-Pay Auctions

02/16/2017 3:10 pm PST

All-pay auctions, a common mechanism for various human and agent interactions, suffers (like many other mechanisms) from the possibility of players' failure to participate in the auction. The authors model such failures and fully characterize equilibrium for this class of games, presenting a symmetric equilibrium and showing that under some conditions the equilibrium is unique. They also reveal various properties of the equilibrium, such as the lack of influence of the most-likely-to-participate player on the behavior of the other players. The authors perform this analysis with two scenarios: the sum-profit model, in which the auctioneer obtains the sum of all submitted bids, and the max-profit model of crowdsourcing contests, in which the auctioneer can only use the best submissions and thus obtains only the winning bid. Finally, the authors examine various methods of influencing the probability of participation such as the effects of misreporting one's own probability of participating and how influencing another player's participation chances changes a player's strategy.

Cooperation and Competition When Bidding for Complex Projects: Centralized and Decentralized Perspectives

02/16/2017 3:05 pm PST

To successfully complete a complex project, agents (companies or individuals) must form a team with the required competencies and resources. A team can be formed either by the project issuer based on individual agents' offers (centralized formation) or by the agents themselves (decentralized formation) bidding for a project as a consortium. The authors investigate rational strategies for agents, propose concepts to characterize the stability of winning teams and study computational complexity of finding these concepts of stability.

Mechanism Design for Demand-Side Management

02/16/2017 3:05 pm PST

As the penetration of renewables into the grid increases, so do the uncertainty and constraints that need to be taken into account during demand-side management (DSM). Mechanism design (MD) provides effective DSM solutions that incorporate end-user preferences and uncertain capabilities, without jeopardizing comfort. The authors discuss the state of the art in DSM methods, which they broadly classify as game theoretic (largely MD) and not. They then proceed to outline a novel scheme for large-scale coordinated demand shifting, a highly important problem. Their mechanism employs the services of cooperatives of electricity consumers, incentivizes truthfulness regarding the contributions promised by the participants, and incorporates profiling techniques that assess contributors' trustworthiness. It is shown via simulations over real-world datasets to effectively shift peak load and generate substantial economic benefits at cooperative and individual levels.

An Algorithm for the Myerson Value in Probabilistic Graphs with an Application to Weighted Voting

02/16/2017 3:07 pm PST

Myerson's graph-restricted games are a well-known formalism for modeling cooperation that's subject to restrictions. In particular, Myerson considered a coalitional game in which cooperation is possible only through an underlying network of links between agents. A unique fair solution concept for graph-restricted games is called the Myerson value. One study generalized these results by considering probabilistic graphs in which agents can cooperate via links only to some extent, that is, with some probability. The authors' algorithm is based on the enumeration of all connected subgraphs in the graph. As a sample application of the new algorithm, they consider a probabilistic graph that represents likelihood of pairwise collaboration between political parties before the 2015 general elections in the UK.

Multiwinner Voting in Genetic Algorithms

02/16/2017 3:06 pm PST

Genetic algorithms are a group of powerful tools for solving ill-posed global optimization problems in continuous domains. When insensitivity in the fitness function is an obstacle, the most desired feature of a genetic algorithm is its ability to explore plateaus of the fitness function surrounding its minimizers. The authors suggest a way of maintaining diversity of the population in the plateau regions based on a new approach for selection according to the theory of multiwinner elections among autonomous agents. The article delivers a detailed description of the new selection algorithm, computational experiments that put the choice of the proper multiwinner rule to use, and a preliminary experiment showing the proposed algorithm's effectiveness in exploring a fitness function's plateau.

Multidefender Security Games

02/16/2017 3:02 pm PST

Current Stackelberg security game models primarily focus on isolated systems in which only one defender is present, despite being part of a more complex system with multiple players. However, many real systems such as transportation networks and the power grid exhibit interdependencies among targets and, consequently, between decision makers jointly charged with protecting them. To understand such multidefender strategic interactions present in security scenarios, the authors investigate security games with multiple defenders. Unlike most prior analyses, they focus on situations in which each defender must protect multiple targets, so even a single defender's best response decision is, in general, nontrivial. Considering interdependencies among targets, the authors develop a novel mixed-integer linear programming formulation to compute a defender's best response, and approximate Nash equilibria of the game using this formulation. Their analysis shows how network structure and the probability of failure spread determine the propensity of defenders to over- or underinvest in security.

Online Prediction via Continuous Artificial Prediction Markets

02/16/2017 3:04 pm PST

Prediction markets are well-established tools for aggregating information from diverse sources into accurate forecasts. Their success has been demonstrated in a wide range applications, including presidential campaigns, sporting events, and economic outcomes. Recently, they've been introduced to the machine learning community in the form of artificial prediction markets, in which algorithms trade contracts reflecting their levels of confidence. To date, these markets have mostly been studied in the context of offline classification, with promising results. The authors extend them to enable their use in online regression and introduce adaptive trading strategies informed by individual trading history and the ability of participants to revise their predictions by reflecting on the wisdom of the crowd, which is manifested in the collective performance of the market. The authors empirically evaluate their model using multiple datasets and show that it outperforms several well-established techniques from the literature on online regression.

An Intelligent System for Spoken Term Detection That Uses Belief Combination

02/16/2017 3:08 pm PST

Spoken term detection (STD) can be considered a subpart of automatic speech recognition that aims to extract partial information from speech signals in the form of query utterances. A variety of STD techniques available in the literature employ a single source of evidence for query utterance match/mismatch determination. In this article, the authors develop an acoustic signal processing-based approach for STD that incorporates several techniques for silence removal, dynamic noise filtration, and evidence combination using the Dempster-Shafer theory. Spectral-temporal features-based voiced segment detection and energy and zero cross rate-based unvoiced segment detection remove silence segments in the speech signal. Comprehensive experiments have been performed on large speech datasets and satisfactory results have been achieved with the proposed approach, which improves existing speaker-dependent STD approaches, specifically the reliability of query utterance spotting, by combining the results from multiple belief sources.

A Quantum-Inspired Feature Fusion Method Based on Maximum Fidelity

02/16/2017 3:11 pm PST

To better reduce redundant data and improve the completeness and conciseness of existing feature data, this articleapplies the theories of fidelity and quantum computation to feature fusion and proposes a novel quantum-inspired method based on maximum fidelity.In contrast tocurrentquantum-inspired feature fusion methods,this method uses the fidelity between feature samples and takes the maximum fidelity and the maximum component of fidelity as key factors to detectand fuse duplicate feature samples in a subset. Fusion results show that the feature fusion method based on maximum fidelity gives better performances regardingrelative completeness and conciseness than current methods and has wide applications in intelligent systems.

Regulating Autonomous Systems: Beyond Standards

02/16/2017 3:06 pm PST

Autonomous systems are on the verge of widespread use, but there are significant challenges to ensuring their safe and appropriate development, deployment, and application. The authors argue that autonomous systems cannot, in general, be regulated using traditional performance standards. They contend that these systems should instead be evaluated by a staged, iterative regulatory approach similar to that for pharmaceuticals and medical devices. This article outlines some features of such a system.

Redecentralizing the Web with Distributed Ledgers

02/16/2017 3:09 pm PST

The web was originally conceived as decentralized and universal, but during its popularization, its big value was built on centralized servers and nonuniversal access. A key element to redecentralize the web is to be able to generate trustable, secure, and accountable updates among autonomous participants without a central server. The authors believe that the marriage between distributed ledgers and linked data can provide this functionality and unlock the web's true potential. As a first step toward it, the authors propose a minimal vocabulary to describe and link distributed ledgers.

An Open Semantic Framework for the Industrial Internet of Things

02/16/2017 3:10 pm PST

This article introduces the Open Semantic Framework (OSF), a one-stop shop for creating and deploying semantic applications as well as their lifecycle management. The authors discuss how the OSF supports knowledge acquisition into semantic knowledge models via novel interface technologies, manages these models internally within core ontologies and pluggable knowledge packs, and provides moderated access to them via a REST API and prefabricated SPARQL queries. As an example of OSF's usage in a practical scenario, the authors show how it can increase worker safety in industrial settings by mitigating workplace safety hazards through the automatic insertion of safety-relevant actions directly in the workflow, based on a model of applicable workplace safety law and regulation. By facilitating the creation, integration, and provisioning of knowledge, the OSF represents an important step on the way to integrating knowledge models worldwide, enabling semantic applications globally to "stand on the shoulders of giants."

Lexicon Generation for Emotion Detection from Text

02/16/2017 3:11 pm PST

General-purpose emotion lexicons (GPELs) that associate words with emotion categories remain a valuable resource for emotion detection. However, the static and formal nature of their vocabularies make them an inadequate resource for detecting emotions in domains that are inherently dynamic in nature. This calls for lexicons that are not only adaptive to the lexical variations in a domain but which also provide finer-grained quantitative estimates to accurately capture word-emotion associations. In this article, the authors demonstrate how to harness labeled emotion text (such as blogs and news headlines) and weakly labeled emotion text (such as tweets) to learn a word-emotion association lexicon by jointly modeling emotionality and neutrality of words using a generative unigram mixture model (UMM). Empirical evaluation confirms that UMM generated emotion language models (topics) have significantly lower perplexity compared to those from state-of-the-art generative models like supervised Latent Dirichlet Allocation (sLDA). Further emotion detection tasks involving word-emotion classification and document-emotion ranking confirm that the UMM lexicon significantly out performs GPELs and also state-of-the-art domain specific lexicons.

A Bright Future for IEEE Intelligent Systems

11/12/2016 2:31 pm PST

Editor-in-chief Daniel Zeng says his farewell as his term ends at the end of 2016.

How to Generate a Good Word Embedding

11/19/2016 2:36 pm PST

The authors analyze three critical components in training word embeddings: model, corpus, and training parameters. They systematize existing neural-network-based word embedding methods and experimentally compare them using the same corpus. They then evaluate each word embedding in three ways: analyzing its semantic properties, using it as a feature for supervised tasks, and using it to initialize neural networks. They also provide several simple guidelines for training good word embeddings.

Can a Machine Replace Humans in Building Regular Expressions? A Case Study

11/19/2016 2:36 pm PST

Regular expressions are routinely used in a variety of different application domains. But building a regular expression involves a considerable amount of skill, expertise, and creativity. In this work, the authors investigate whether a machine can surrogate these qualities and automatically construct regular expressions for tasks of realistic complexity. They discuss a large-scale experiment involving more than 1,700 users on 10 challenging tasks. The authors compare the solutions constructed by these users to those constructed by a tool based on genetic programming that they recently developed and made publicly available. The quality of automatically constructed solutions turned out to be similar to the quality of those constructed by the most skilled user group; the time for automatic construction was likewise similar to the time required by human users.

Influence of Rating Prediction on Group Recommendation's Accuracy

11/19/2016 2:37 pm PST

Recommender systems suggest items that might be interesting to a user. To achieve this, rating prediction is the main form of information processing that these systems perform. This article tackles the problem of predicting ratings in a group recommender system by analyzing how system accuracy is influenced by the choice of prediction approach and by a solution that employs the predicted values to avoid data sparsity. The results of more than 100 experiments show that by predicting the ratings for individual users instead of predicting them for groups, and by using these predictions in a system's group detection task, accuracy increases and problems caused by data sparsity are reduced.