A large variety of issues influence the success of data mining on a given problem. Two primary and important issues are the representation and the quality of the dataset. Specifically, if much redundant and unrelated or noisy and unreliable information is presented, then knowledge discovery becomes a very difficult problem. It is well-known that data preparation steps require significant processing time in machine learning tasks. It would be very helpful and quite useful if there were various preprocessing algorithms with the same reliable and effective performance across all datasets, but this is impossible. To this end, we present the most well-known and widely used up-to-date algorithms for each step of data preprocessing in the framework of predictive data mining.

Research on Multi-Agent Systems (MAS) has led to the development of several models, languages, and technologies for programming not only agents, but also their interaction, the application environment where they are situated, as well as the organization in which they participate. Research on those topics moved from agent-oriented programming towards multi-agent-oriented programming (MAOP). A MAS program is then designed and developed using a structured set of concepts and associated first-class design and programming abstractions that go beyond the concepts normally associated with agents. They include those related to environment, interaction, and organization. JaCaMo is a platform for MAOP built on top of three seamlessly integrated dimensions (i.e. structured sets of concepts and associated execution platforms): for programming belief desire intention (BDI) agents, their artefact-based environments, and their normative organizations. The key purpose of our work on JaCaMo is to support programmers in exploring the synergy between these dimensions, providing a comprehensive programming model, as well as a corresponding platform for developing and running MAS. This paper provides a practical overview of MAOP using JaCaMo. We show how emphasizing one particular dimension leads to different solutions to the same problem, and discuss the issues of each of those solutions.

Recently, valuable knowledge that can be retrieved from a huge volume of datasets (called Big Data) set in motion the development of frameworks to process data based on parallel and distributed computing, including Apache Hadoop, Facebook Corona, and Microsoft Dryad. Apache Hadoop is an open source implementation of Google MapReduce that attracted strong attention from the research community both in academia and industry. Hadoop MapReduce scheduling algorithms play a critical role in the management of large commodity clusters, controlling QoS requirements by supervising users, jobs, and tasks execution. Hadoop MapReduce comprises three schedulers: FIFO, Fair, and Capacity. However, the research community has developed new optimizations to consider advances and dynamic changes in hardware and operating environments. Numerous efforts have been made in the literature to address issues of network congestion, straggling, data locality, heterogeneity, resource under-utilization, and skew mitigation in Hadoop scheduling. Recently, the volume of research published in journals and conferences about Hadoop scheduling has consistently increased, which makes it difficult for researchers to grasp the overall view of research and areas that require further investigation. A scientific literature review has been conducted in this study to assess preceding research contributions to the Apache Hadoop scheduling mechanism. We classify and quantify the main issues addressed in the literature based on their jargon and areas addressed. Moreover, we explain and discuss the various challenges and open issue aspects in Hadoop scheduling optimizations.

Generalized planning studies the representation, computation and evaluation of solutions that are valid for multiple planning instances. These are topics studied since the early days of AI. However, in recent years, we are experiencing the appearance of novel formalisms to compactly represent generalized planning tasks, the solutions to these tasks (called generalized plans) and efficient algorithms to compute generalized plans. The paper reviews recent advances in generalized planning and relates them to existing planning formalisms, such as planning with domain control knowledge and approaches for planning under uncertainty, that also aim at generality.

In this paper, we describe how predictive models can be positively exploited in abstract argumentation. In particular, we present two main sets of results. On one side, we show that predictive models are effective for performing algorithm selection in order to determine which approach is better to enumerate the preferred extensions of a given argumentation framework. On the other side, we show that predictive models predict significant aspects of the solution to the preferred extensions enumeration problem. By exploiting an extensive set of argumentation framework features—that is, values that summarize a potentially important property of a framework—the proposed approach is able to provide an accurate prediction about which algorithm would be faster on a given problem instance, as well as of the structure of the solution, where the complete knowledge of such structure would require a computationally hard problem to be solved. Improving the ability of existing argumentation-based systems to support human sense-making and decision processes is just one of the possible exploitations of such knowledge obtained in an inexpensive way.

Social networks occupy a ubiquitous and pervasive place in the life of their users. The substantial amount of content generated and shared by social networking users offers new research opportunities across a wide variety of disciplines, including media and communication studies, linguistics, sociology, psychology, information and computer sciences, or education. This situation, in combination with the continuous growth of social media data, creates an imperative need for content organisation. Thus, large-scale text learning tasks in social environments arise as one of the most relevant problems in machine learning and data mining. Interestingly, social media data pose several challenges due to its sparse, high-dimensional and large-volume characteristics. This survey reviews the field of social media data learning, focusing on classification and clustering techniques, as they are two of the most frequent learning tasks. It reviews not only new techniques that have been developed to tackle the new challenges posed by short-texts, but also how traditional techniques can be adapted to overcome such challenges. Then, open issues and research opportunities for social media data learning are discussed.

One challenge faced by reinforcement learning (RL) agents is that in many environments the reward signal is sparse, leading to slow improvement of the agent’s performance in early learning episodes. Potential-based reward shaping can help to resolve the aforementioned issue of sparse reward by incorporating an expert’s domain knowledge into the learning through a potential function. Past work on reinforcement learning from demonstration (RLfD) directly mapped (sub-optimal) human expert demonstration to a potential function, which can speed up RL. In this paper we propose an introspective RL agent that significantly further speeds up the learning. An introspective RL agent records its state–action decisions and experience during learning in a priority queue. Good quality decisions, according to a Monte Carlo estimation, will be kept in the queue, while poorer decisions will be rejected. The queue is then used as demonstration to speed up RL via reward shaping. A human expert’s demonstration can be used to initialize the priority queue before the learning process starts. Experimental validation in the 4-dimensional CartPole domain and the 27-dimensional Super Mario AI domain shows that our approach significantly outperforms non-introspective RL and state-of-the-art approaches in RLfD in both domains.

An increase in volumes of data and a shift towards live data enabled a stronger focus on resource-intensive tasks which run continuously over long periods. A Grid has potential to offer the required resources for these tasks, while considering a fair and balanced allocation of resources among multiple client agents. Taking this into account, a Grid might be unwilling to allocate its resources for long time, leading to task interruptions. This problem becomes even more serious if an interruption of one task may lead to the interruption of dependent tasks. Here, we discuss a new strategy for resource re-allocation which is utilized by a client with the aim to prevent too long interruptions by re-allocating resources between its own tasks. Those re-allocations are suggested by a client agent, but only a Grid can re-allocate resources if agreed. Our strategy was tested under the different Grid settings, accounting for the adjusted coefficients, and demonstrated noticeable improvements in client utilities as compared to when it is not considered. Our experiment was also extended to tests with environmental modelling and realistic Grid resource simulation, grounded in real-life Grid studies. These tests have also shown a useful application of our strategy.

Deep reinforcement learning (deep RL) has achieved superior performance in complex sequential tasks by using deep neural networks as function approximators to learn directly from raw input images. However, learning directly from raw images is data inefficient. The agent must learn feature representation of complex states in addition to learning a policy. As a result, deep RL typically suffers from slow learning speeds and often requires a prohibitively large amount of training time and data to reach reasonable performance, making it inapplicable to real-world settings where data are expensive. In this work, we improve data efficiency in deep RL by addressing one of the two learning goals, feature learning. We leverage supervised learning to pre-train on a small set of non-expert human demonstrations and empirically evaluate our approach using the asynchronous advantage actor-critic algorithms in the Atari domain. Our results show significant improvements in learning speed, even when the provided demonstration is noisy and of low quality.

In the reinforcement learning context, a landmark is a compact information which uniquely couples a state, for problems with hidden states. Landmarks are shown to support finding good memoryless policies for Partially Observable Markov Decision Processes (POMDP) which contain at least one landmark. SarsaLandmark, as an adaptation of Sarsa(λ), is known to promise a better learning performance with the assumption that all landmarks of the problem are known in advance.

In this paper, we propose a framework built upon SarsaLandmark, which is able to automatically identify landmarks within the problem during learning without sacrificing quality, and requiring no prior information about the problem structure. For this purpose, the framework fuses SarsaLandmark with a well-known multiple-instance learning algorithm, namely Diverse Density (DD). By further experimentation, we also provide a deeper insight into our concept filtering heuristic to accelerate DD, abbreviated as DDCF (Diverse Density with Concept Filtering), which proves itself to be suitable for POMDPs with landmarks. DDCF outperforms its antecedent in terms of computation speed and solution quality without loss of generality.

The methods are empirically shown to be effective via extensive experimentation on a number of known and newly introduced problems with hidden state, and the results are discussed.

Among an array of techniques proposed to speed-up reinforcement learning (RL), learning from human demonstration has a proven record of success. A related technique, called Human-Agent Transfer, and its confidence-based derivatives have been successfully applied to single-agent RL. This article investigates their application to collaborative multi-agent RL problems. We show that a first-cut extension may leave room for improvement in some domains, and propose a new algorithm called coordination confidence (CC). CC analyzes the difference in perspectives between a human demonstrator (global view) and the learning agents (local view) and informs the agents’ action choices when the difference is critical and simply following the human demonstration can lead to miscoordination. We conduct experiments in three domains to investigate the performance of CC in comparison with relevant baselines.

In order to enable robots to interact with humans in a natural way, they need to be able to autonomously learn new tasks. The most natural way for humans to tell another agent, which can be a human or robot, to perform a task is via natural language. Thus, natural human–robot interactions also require robots to understand natural language, i.e. extract the meaning of words and phrases. To do this, words and phrases need to be linked to their corresponding percepts through grounding. Afterward, agents can learn the optimal micro-action patterns to reach the goal states of the desired tasks. Most previous studies investigated only learning of actions or grounding of words, but not both. Additionally, they often used only a small set of tasks as well as very short and unnaturally simplified utterances. In this paper, we introduce a framework that uses reinforcement learning to learn actions for several tasks and cross-situational learning to ground actions, object shapes and colors, and prepositions. The proposed framework is evaluated through a simulated interaction experiment between a human tutor and a robot. The results show that the employed framework can be used for both action learning and grounding.

One way to address this low sample efficiency of reinforcement learning (RL) is to employ human expert demonstrations to speed up the RL process (RL from demonstration or RLfD). The research so far has focused on demonstrations from a single expert. However, little attention has been given to the case where demonstrations are collected from multiple experts, whose expertise may vary on different aspects of the task. In such scenarios, it is likely that the demonstrations will contain conflicting advice in many parts of the state space. We propose a two-level Q-learning algorithm, in which the RL agent not only learns the policy of deciding on the optimal action but also learns to select the most trustworthy expert according to the current state. Thus, our approach removes the traditional assumption that demonstrations come from one single source and are mostly conflict-free. We evaluate our technique on three different domains and the results show that the state-of-the-art RLfD baseline fails to converge or performs similarly to conventional Q-learning. In contrast, the performance level of our novel algorithm increases with more experts being involved in the learning process and the proposed approach has the capability to handle demonstration conflicts well.

User validation is one of the challenges facing the ontology alignment community, as there are limits to the quality of the alignments produced by automated alignment algorithms. In this paper, we present a broad study on user validation of ontology alignments that encompasses three distinct but inter-related aspects: the profile of the user, the services of the alignment system, and its user interface. We discuss key issues pertaining to the alignment validation process under each of these aspects and provide an overview of how current systems address them. Finally, we use experiments from the Interactive Matching track of the Ontology Alignment Evaluation Initiative 2015–2018 to assess the impact of errors in alignment validation, and how systems cope with them as function of their services.

In this study, we designed a localization and obstacle avoidance system for humanoid robots in the Federation of International Robot-soccer Association (FIRA) HuroCup united soccer competition event. The localization is implemented by using grid points, gait, and steps to determine the positions of each robot. To increase the localization accuracy and eliminate the accumulated distance errors resulting from step counting, the localization is augmented with image pattern matching using a system model. The system also enables the robot to determine the ball’s position on the field using a color model of the ball. Moreover, to avoid obstacles, the robots calculate the obstacle distance using data extracted from real-time images and determine a suitable direction for movement. With the integration of this accurate self-localization algorithm, ball identification scheme, and obstacle avoidance system, the robot team is capable of accomplishing the necessary tasks for a FIRA soccer game.

The current fourth industrial revolution, or ‘Industry 4.0’ (I4.0), is driven by digital data, connectivity, and cyber systems, and it has the potential to create impressive/new business opportunities. With the arrival of I4.0, the scenario of various intelligent systems interacting reliably and securely with each other becomes a reality which technical systems need to address. One major aspect of I4.0 is to adopt a coherent approach for the semantic communication in between multiple intelligent systems, which include human and artificial (software or hardware) agents. For this purpose, ontologies can provide the solution by formalizing the smart manufacturing knowledge in an interoperable way. Hence, this paper presents the few existing ontologies for I4.0, along with the current state of the standardization effort in the factory 4.0 domain and examples of real-world scenarios for I4.0.

This paper presents our sketch drawing artist humanoid robot research. One of the limitations of the existing artist humanoid robot is the lack of feedback on the error that occurs during the drawing process. The contribution of this research is the development of a humanoid robot artist with drawing error correction capability. Based on our previous work with open-loop control pen-and-ink humanoid robot artist, we have implemented a closed-loop visual servoing approach to address this problem. Our experimental results show that this approach is sufficient to correct drawing errors that occur due to mechanical limitation of a robot.

The team-orienteering problem (TOP) has broad applicability. Examples of possible uses are in factory and automation settings, robot sports teams, and urban search and rescue applications. We chose the rescue domain as a guiding example throughout this paper. Hence, this paper explores a practical variant of TOP with time window (TOPTW) for rescue applications by humanoid robots called TOPTWR. Due to the significant range of algorithm choices and their parameters tuning challenges, the use of hyper-heuristics is recommended. Hyper-heuristics can select, order, or generate different low-level heuristics with different optimization algorithms. In this paper, first, a general multi-objective (MO) solution is defined, with five objectives for TOPTWR. Then a robust and efficient MO and evolutionary hyper-heuristic algorithm for TOPTW based on the humanoid robot’s characteristics in the rescue applications (MOHH-TOPTWR) is proposed. MOHH-TOPTWR includes two MO evolutionary metaheuristics algorithms (MOEAs) known as non-dominated sorting genetic algorithm (NSGA-III) and MOEA based on decomposition (MOEA/D). In this paper, new benchmark instances are proposed for rescue applications using the existing ones for TOPTW. The experimental results show that MOHH-TOPTWR in both MOEAs can outperform all the state-of-the-art algorithms as well as NSGA-III and MOEA/D MOEAs.

The development of a versatile, fully-capable humanoid robot as envisioned in science fiction books is one of the most challenging but interesting issues in the robotic field. Currently, existing humanoid robots are designed with different purposes and applications in mind. In humanoid robot development process, each robot is designed with various characteristics, abilities, and equipment, which influence the general structure, cost, and difficulty of development. Even though humanoid robot development is very popular, a few review papers are focusing on the design and development process of humanoid robots. Motivated by this, we present this review paper to show variations in the requirements, design, and development process and also propose a taxonomy of existing humanoid robots. It aims at demonstrating a general perspective of existing humanoid robots’ characteristics and applications. This paper includes state-of-the-art and successfully reported existing humanoid robot designs along with different robots used in various robot competitions.