International Journal on Artificial Intelligence Tools

This paper is based on my Keynote Address at the Tools with AI conference on November 4, 2013. The talk focussed on some factors used to determine a user's context — those attributes, both tacit and explicit, which help to ascertain a user's intensions for a search request in order to make a decision. I also explored how social semantic search can assist in guiding the decision process, especially when the decision is of a personal nature, for example, in decisions involving health care, where there may be a number of avenues to pursue. I focussed on our patented-technology exemplified by the Knowledge Sifter system, and an application called Personal Health Explorer (PHE), a semantic recommender system that allows an individual to perform semantic search and discovery related to conditions and diseases contained in his personal health record (PHR). The PHE system consults authoritative ontologies and reputable information sources to provide semantically enhanced authoritative recommendations that can be stored in the individual's (PHR) for further research and consultation. The PHE is implemented using the Knowledge Sifter agent-based framework and Microsoft's HealthVault.

Real-time autonomy is a key element for system which closes the loop between observation, interpretation, planning, and action, commonly found in UxV, robotics, smart vehicle technologies, automated industrial machineries, and autonomic computing. Real-time autonomic cyber system requires timely and accurate decision making and adaptive planning. Autonomic decision making understands its own state and the perceived state of its environment. It is capable of anticipating changes and future states and projecting the effects of actions into future states. Understanding of current state and the knowledge/model of the world are needed for extrapolating actions and deriving action plans. This position paper proposes a hybrid, statistical-formal approach toward achieving realtime autonomy.

This survey gives a review of recent artificial intelligence-related research directions that are considered priority areas by the U.S. Air Force and targeted for basic research funding by Air Force Office of Scientific Research. These research areas include space situational awareness, autonomous systems, sensing and information fusion, surveillance, navigation, robust decision making, human-computer interfaces, and computational and machine intelligence. The possible contributions of artificial intelligence to these topics will be described and illustrated whenever possible by recently awarded grants.

Pattern mining provides useful tools for exploratory data analysis. Numerous efficient algorithms exist that are able to discover various types of patterns in large datasets. Unfortunately, the problem of identifying patterns that are genuinely interesting to a particular user remains challenging. Current approaches generally require considerable data mining expertise or effort from the data analyst, and hence cannot be used by typical domain experts.

To address this, we introduce a generic framework for interactive learning of userspecific pattern ranking functions. The user is only asked to rank small sets of patterns, while a ranking function is inferred from this feedback by preference learning techniques. Moreover, we propose a number of active learning heuristics to minimize the effort required from the user, while ensuring that accurate rankings are obtained. We show how the learned ranking functions can be used to mine new, more interesting patterns.

We demonstrate two concrete instances of our framework for two different pattern mining tasks, frequent itemset mining and subgroup discovery. We empirically evaluate the capacity of the algorithm to learn pattern rankings by emulating users. Experiments demonstrate that the system is able to learn accurate rankings, and that the active learning heuristics help reduce the required user effort. Furthermore, using the learned ranking functions as search heuristics allows discovering patterns of higher quality than those in the initial set. This shows that machine learning techniques in general, and active preference learning in particular, are promising building blocks for interactive data mining systems.

A scalable, reactive and easy to evolve reasoning mechanism is essential for the success of automated smart environments, augmented with a large number of sensors and actuators. While constraint satisfaction problem (CSP) model is applicable for modelling decision making in such environments, the straightforward representation of the model as a CSP leads to a great number of excessive calculations. In this paper, we propose a method of modelling the task as a Dynamic CSP in a way that avoids unnecessary recalculations with new events in the environment. We present a Dependency Graph data structure, which not only allows to reduce CSP search space for every consecutive sensor event by detecting only affected parts of the environment, but also allows to give enough information to users of the system to specify the exact reasons of system's decisions, even with a large number of constraints. We formally prove that partial recalculation of affected parts still keeps the full environment globally satisfied and globally optimal. The evaluation of the system in the living lab showed real-time responses for all events. Additional simulated performance experiments showed that the Dependency Graph approach consistently outperforms the straightforward CSP representation. The experiments also showed that the clusterization of the environment has a noticeable effect on the performance, with highly clusterized environments requiring less computations.

Recently, the state-of-the-art AI planners have significantly improved planning efficiency on Fully Observable Nondeterministic planning (FOND) problems with strong cyclic solutions. These strong cyclic solutions are guaranteed to achieve the goal if they terminate, implying that there is a possibility that they may run into indefinite loops. In contrast, strong solutions are guaranteed to achieve the goal, but few planners can effectively handle FOND problems with strong solutions. In this study, we aim to address this difficult, yet under-investigated class of planning problems: FOND planning problems with strong solutions. We present a planner that employs a new data structure, MRDAG (multi-root directed acyclic graph), to define how the solution space should be expanded. Based on the characteristics of MRDAG, we develop heuristics to ensure planning towards the relevant search direction and design optimizations to prune the search space to further improve planning efficiency. We perform extensive experiments to evaluate MRDAG, the heuristics, and the optimizations for pruning the search space. Experimental results show that our strong algorithm achieves impressive performance on a variety of benchmark problems: on average it runs more than three orders of magnitude faster than the state-of-the-art planners, MBP and Gamer, while demonstrating significantly better scalability.

The shape or contour of an object is usually stable and persistent, so it is a good basis for invariant recognition. For this purpose, two problems must be handled. The first is obtaining clean edges and the other is organizing those edges into a structured form so that they can be manipulated easily. We apply a bio-inspired orientation detection algorithm because it can output a fairly clean set of lines, and all lines are in the form of vectors instead of pixels. This line representation is efficient. We decompose them into several slope-depended layers and then create a hierarchical partition tree to record their geometric distribution. Based on the similarity of trees, a rough classification of objects can be realized. However, for an accuracy recognition, we design a moment-based measure to describe the detail layout of lines in a layer and then re-describe image by Hu's moment invariants. The experimental results suggest that the representation efficiency enabled by simple cell's neural mechanism and application of multi-layered representation schema can simplify the complexity of the algorithm. This proves that line-context representation greatly eases subsequent shape-oriented recognition.

This work focuses on learning the structure of Markov networks from data. Markov networks are parametric models for compactly representing complex probability distributions. These models are composed by: a structure and numerical weights, where the structure describes independences that hold in the distribution. Depending on which is the goal of structure learning, learning algorithms can be divided into: density estimation algorithms, where structure is learned for answering inference queries; and knowledge discovery algorithms, where structure is learned for describing independences qualitatively. The latter algorithms present an important limitation for describing independences because they use a single graph; a coarse grain structure representation which cannot represent flexible independences. For instance, context-specific independences cannot be described by a single graph. To overcome this limitation, this work proposes a new alternative representation named canonical model as well as the CSPC algorithm; a novel knowledge discovery algorithm for learning canonical models by using context-specific independences as constraints. On an extensive empirical evaluation, CSPC learns more accurate structures than state-of-the-art density estimation and knowledge discovery algorithms. Moreover, for answering inference queries, our approach obtains competitive results against density estimation algorithms, significantly outperforming knowledge discovery algorithms.

Social recommender systems exploit social knowledge available in social networks to provide accurate recommendations. However, their instantiation is not straightforward due to its complexity. To alleviate this development complexity, we propose a methodology based on templates that conceptualize the behavior of such applications and can be reused to create several social recommender applications in social networks. This development methodology comprises not only templates but also a generic architecture named arise and a collection of software components that provide the required functionality. We prove that our social templates speed up and facilitate the development process, and demonstrate the viability of our generic architecture in two different case studies.

Despite the advances made in the last decade in automated planning, no planner outperforms all the others in every known benchmark domain. This observation motivates the idea of selecting different planning algorithms for different domains. Moreover, the planners' performances are affected by the structure of the search space, which depends on the encoding of the considered domain. In many domains, the performance of a planner can be improved by exploiting additional knowledge, for instance, in the form of macro-operators or entanglements.

In this paper we propose ASAP, an automatic Algorithm Selection Approach for Planning that: (i) for a given domain initially learns additional knowledge, in the form of macro-operators and entanglements, which is used for creating different encodings of the given planning domain and problems, and (ii) explores the 2 dimensional space of available algorithms, defined as encodings–planners couples, and then (iii) selects the most promising algorithm for optimising either the runtimes or the quality of the solution plans.

The spiral has universal appeal and a mysterious resonance with the human spirit. It is intriguing and beautiful. The spiral pattern is found extensively in nature, encoded into plants, animals, weather phenomena, the earth, and galaxies around us. The paper starts with a mathematical representation of the spiral phenomena. Next, 2-dimensional and n-dimensional Spiral Dynamics Optimization (SpDO) algorithms are presented followed by their variants such as adaptive SpDO and hybrid SpDO. Finally, applications to engineering problems are reviewed including a flexible robot manipulator system, training of artificial neural networks, the Combined Economic and Emission Dispatch (CEED) optimization problem in power system operations, and clustering in data mining.

This article presents a review of a novel nature-inspired adaptive optimization algorithm known as Water Drop Algorithm (WDA) which imitates the dynamics of river systems and the behavior of water drops when they are moving in a river system such as the variation of velocity, the change of sediment/soil in the river bed, the change of direction of the flow, and so on. Recent improved, modified, and adaptive WDAs are examined in this paper. Various applications of WDA are described briefly including the vehicle routing problem, economic load dispatch problem, economic and emission dispatch problem, optimal data aggregation tree in wireless networks, reactive power dispatch problem service selection, and reservoir operation.