MMA: a multi-view and multi-modality benchmark dataset for human action recognition

Human action recognition is an active research topic in both computer vision and machine learning communities, which has broad applications including surveillance, biometrics and human computer interaction. In the past decades, although some famous action datasets have been released, there still exist limitations, including the limited action categories and samples, camera views and variety of scenarios. Moreover, most of them are designed for a subset of the learning problems, such as single-view learning problem, cross-view learning problem and multi-task learning problem. In this paper, we introduce a multi-view, multi-modality benchmark dataset for human action recognition (abbreviated to MMA). MMA consists of 7080 action samples from 25 action categories, including 15 single-subject actions and 10 double-subject interactive actions in three views of two different scenarios. Further, we systematically benchmark the state-of-the-art approaches on MMA with respective to all three learning problems by different temporal-spatial feature representations. Experimental results demonstrate that MMA is challenging on all three learning problems due to significant intra-class variations, occlusion issues, views and scene variations, and multiple similar action categories. Meanwhile, we provide the baseline for the evaluation of existing state-of-the-art algorithms.     

Ranking policies in discrete Markov decision processes

An optimal probabilistic-planning algorithm solves a problem, usually modeled by a Markov decision process, by finding an optimal policy. In this paper, we study the k best policies problem. The problem is to find the k best policies of a discrete Markov decision process. The k best policies, k?>?1, cannot be found directly using dynamic programming. Naïvely, finding the k-th best policy can be Turing reduced to the optimal planning problem, but the number of problems queried in the naïve algorithm is exponential in k. We show empirically that solving k best policies problem by using this reduction requires unreasonable amounts of time even when k?=?3. We then provide two new algorithms. The first is a complete algorithm, based on our theoretical contribution that the k-th best policy differs from the i-th policy, for some i?k, on exactly one state. The second is an approximate algorithm that skips many less useful policies. We show that both algorithms have good scalability. We also show that the approximate algorithms runs much faster and finds interesting, high-quality policies.     

A memetic algorithm for the capacitated m-ring-star problem

The Capacitated m-Ring-Star Problem (CmRSP) models a network topology design problem in the telecommunications industry. In this paper, we propose to solve this problem using a memetic algorithm that includes a crossover operation, a mutation operation, a local search involving three neighborhood operators, and a population selection strategy that maintains population diversity. Our approach generates the best known solutions for 131 out of 138 benchmark instances, improving on the previous best solutions for 24 of them, and exhibits more advantages on large benchmark instances when compared with the best existing approach. Additionally, all existing algorithms for this problem in literature assume that the underlying graph of the problem instance satisfies the triangle inequality rule; our approach does not require this assumption. We also generated a new set of 36 larger test instances based on a digital data service network price structure to serve as a new benchmark data set for future researchers.     

Extracting Actionable Knowledge from Decision Trees

Most data mining algorithms and tools stop at discovered customer models, producing distribution information on customer profiles. Such techniques, when applied to industrial problems such as customer relationship management (CRM), are useful in pointing out customers who are likely attritors and customers who are loyal, but they require human experts to postprocess the discovered knowledge manually. Most of the postprocessing techniques have been limited to producing visualization results and interestingness ranking, but they do not directly suggest actions that would lead to an increase in the objective function such as profit. In this paper, we present novel algorithms that suggest actions to change customers from an undesired status (such as attritors) to a desired one (such as loyal) while maximizing an objective function: the expected net profit. These algorithms can discover cost-effective actions to transform customers from undesirable classes to desirable ones. The approach we take integrates data mining and decision making tightly by formulating the decision making problems directly on top of the data mining results in a postprocessing step. To improve the effectiveness of the approach, we also present an ensemble of decision trees which is shown to be more robust when the training data changes. Empirical tests are conducted on both a realistic insurance application domain and UCI benchmark data     

Model, solution concept, and Kth-best algorithm for linear trilevel programming

Trilevel programming refers to hierarchical optimization problems in which the top-level, middle-level, and bottom-level decision entities all attempt to optimize their individual objectives, but are impacted by the actions and partial control exercised by decision entities located at other levels. To solve this complex problem, in this study first we propose the use of a general linear trilevel programming (LTLP) subsequently, we develop a trilevel Kth-best algorithm to solve LTLP problems. A user-friendly trilevel decision support tool is also developed. A case study further illustrates the effectiveness of the proposed method.     

A <Emphasis Type="Italic">Q</Emphasis>-learning-based swarm optimization algorithm for economic dispatch problem

In this paper, we treat optimization problems as a kind of reinforcement learning problems regarding an optimization procedure for searching an optimal solution as a reinforcement learning procedure for finding the best policy to maximize the expected rewards. This viewpoint motivated us to propose a Q-learning-based swarm optimization (QSO) algorithm. The proposed QSO algorithm is a population-based optimization algorithm which integrates the essential properties of Q-learning and particle swarm optimization. The optimization procedure of the QSO algorithm proceeds as each individual imitates the behavior of the global best one in the swarm. The best individual is chosen based on its accumulated performance instead of its momentary performance at each evaluation. Two data sets including a set of benchmark functions and a real-world problem—the economic dispatch (ED) problem for power systems—were used to test the performance of the proposed QSO algorithm. The simulation results on the benchmark functions show that the proposed QSO algorithm is comparable to or even outperforms several existing optimization algorithms. As for the ED problem, the proposed QSO algorithm has found solutions better than all previously found solutions.     

Learning from positive and unlabeled examples

In many machine learning settings, labeled examples are difficult to collect while unlabeled data are abundant. Also, for some binary classification problems, positive examples which are elements of the target concept are available. Can these additional data be used to improve accuracy of supervised learning algorithms? We investigate in this paper the design of learning algorithms from positive and unlabeled data only. Many machine learning and data mining algorithms, such as decision tree induction algorithms and naive Bayes algorithms, use examples only to evaluate statistical queries (SQ-like algorithms). Kearns designed the statistical query learning model in order to describe these algorithms. Here, we design an algorithm scheme which transforms any SQ-like algorithm into an algorithm based on positive statistical queries (estimate for probabilities over the set of positive instances) and instance statistical queries (estimate for probabilities over the instance space). We prove that any class learnable in the statistical query learning model is learnable from positive statistical queries and instance statistical queries only if a lower bound on the weight of any target concept f can be estimated in polynomial time. Then, we design a decision tree induction algorithm POSC4.5, based on C4.5, that uses only positive and unlabeled examples and we give experimental results for this algorithm. In the case of imbalanced classes in the sense that one of the two classes (say the positive class) is heavily underrepresented compared to the other class, the learning problem remains open. This problem is challenging because it is encountered in many real-world applications.     

Arbitrary function optimisation with metaheuristics

No free lunch theorems for optimisation suggest that empirical studies on benchmarking problems are pointless, or even cast negative doubts, when algorithms are being applied to other problems not clearly related to the previous ones. Roughly speaking, reported empirical results are not just the result of algorithms’ performances, but the benchmark used therein as well; and consequently, recommending one algorithm over another for solving a new problem might be always disputable. In this work, we propose an empirical framework, arbitrary function optimisation framework, that allows researchers to formulate conclusions independent of the benchmark problems that were actually addressed, as long as the context of the problem class is mentioned. Experiments on sufficiently general scenarios are reported with the aim of assessing this independence. Additionally, this article presents, to the best of our knowledge, the first thorough empirical study on the no free lunch theorems, which is possible thanks to the application of the proposed methodology, and whose main result is that no free lunch theorems unlikely hold on the set of binary real-world problems. In particular, it is shown that exploiting reasonable heuristics becomes more beneficial than random search when dealing with binary real-world applications.     

The complexity of agent design problems: Determinism and history dependence

The agent design problem is as follows: given a specification of an environment, together with a specification of a task, is it possible to construct an agent that can be guaranteed to successfully accomplish the task in the environment? In this article, we study the computational complexity of the agent design problem for tasks that are of the form “achieve this state of affairs” or “maintain this state of affairs.” We consider three general formulations of these problems (in both non-deterministic and deterministic environments) that differ in the nature of what is viewed as an “acceptable” solution: in the least restrictive formulation, no limit is placed on the number of actions an agent is allowed to perform in attempting to meet the requirements of its specified task. We show that the resulting decision problems are intractable, in the sense that these are non-recursive (but recursively enumerable) for achievement tasks, and non-recursively enumerable for maintenance tasks. In the second formulation, the decision problem addresses the existence of agents that have satisfied their specified task within some given number of actions. Even in this more restrictive setting the resulting decision problems are either pspace-complete or np-complete. Our final formulation requires the environment to be history independent and bounded. In these cases polynomial time algorithms exist: for deterministic environments the decision problems are nl-complete; in non-deterministic environments, p-complete.     

An Experimental Study of Algorithms for Weighted Completion Time Scheduling

We consider the total weighted completion time scheduling problem for parallel identical machines and precedence constraints, P| prec|\sum w _i C _i . This important and broad class of problems is known to be NP-hard, even for restricted special cases, and the best known approximation algorithms have worst-case performance that is far from optimal. However, little is known about the experimental behavior of algorithms for the general problem. This paper represents the first attempt to describe and evaluate comprehensively a range of weighted completion time scheduling algorithms. We first describe a family of combinatorial scheduling algorithms that optimally solve the single-machine problem, and show that they can be used to achieve good performance for the multiple-machine problem. These algorithms are efficient and find schedules that are on average within 1.5\percent of optimal over a large synthetic benchmark consisting of trees, chains, and instances with no precedence constraints. We then present several ways to create feasible schedules from nonintegral solutions to a new linear programming relaxation for the multiple-machine problem. The best of these linear programming-based approaches finds schedules that are within 0.2\percent of optimal over our benchmark. Finally, we describe how the scheduling phase in profile-based program compilation can be expressed as a weighted completion time scheduling problem and apply our algorithms to a set of instances extracted from the SPECint95 compiler benchmark. For these instances with arbitrary precedence constraints, the best linear programming-based approach finds optimal solutions in 78\percent of cases. Our results demonstrate that careful experimentation can help lead the way to high quality algorithms, even for difficult optimization problems. Received October 30, 1998; revised March 28, 2001.     

Efficient algorithms for consensus string problems minimizing both distance sum and radius

The consensus (string) problem is finding a representative string, called a consensus, of a given set S of strings. In this paper we deal with consensus problems considering both distance sum and radius, where the distance sum is the sum of (Hamming) distances from the strings in S to the consensus and the radius is the longest (Hamming) distance from the strings in S to the consensus. Although there have been results considering either distance sum or radius, there have been no results considering both, to the best of our knowledge.We present the first algorithms for two consensus problems considering both distance sum and radius for three strings: one problem is to find an optimal consensus minimizing both distance sum and radius. The other problem is to find a bounded consensus such that the distance sum is at most s and the radius is at most r for given constants s and r. Our algorithms are based on characterization of the lower bounds of distance sum and radius, and thus they solve the problems efficiently. Both algorithms run in linear time.     

Metaheuristics based on decision hierarchies for the traveling purchaser problem

In this paper we address the traveling purchaser problem, an NP‐hard problem that generalizes the traveling salesman problem. We present several metaheuristics that combine genetic algorithms and local search. The genetic algorithms are induced by different hierarchic orderings of the decision making regarding the route and the acquisition of the items. Computational experiments were carried out with benchmark instances and the results show that the proposed metaheuristics are a suitable tool to solve high‐dimensioned instances for which the exact methods do not provide solutions within a reasonable CPU time. For several instances, best new upper bounds for the optimum value of the objective function were obtained.     

A discrete particle swarm optimization algorithm with self-adaptive diversity control for the permutation flowshop problem with blocking

This paper proposes a discrete particle swarm optimization (DPSO) algorithm for the m-machine permutation flowshop scheduling problem with blocking to minimize the makespan, which has a strong industrial background, e.g., many production processes of chemicals and pharmaceuticals in chemical industry can be reduced to this problem. To prevent the DPSO from premature convergence, a self-adaptive diversity control strategy is adopted to diversify the population when necessary by adding a random perturbation to the velocity of each particle according to a probability controlled by the diversity of the current population. In addition, a stochastic variable neighborhood search is used as the local search to improve the search intensification. Computational results using benchmark problems show that the proposed DPSO algorithm outperforms previous algorithms proposed in the literature and that it can obtain 111 new best known upper bounds for the 120 benchmark problems.     

Action dependencies in privacy-preserving multi-agent planning

Collaborative privacy-preserving planning (CPPP) is a multi-agent planning task in which agents need to achieve a common set of goals without revealing certain private information. In many CPPP algorithms, the individual agents reason about a projection of the multi-agent problem onto a single-agent classical planning problem. For example, an agent can plan as if it controls the public actions of other agents, ignoring any private preconditions and effects theses actions may have, and use the cost of this plan as a heuristic estimate of the cost of the full, multi-agent plan. Using such a projection, however, ignores some dependencies between agents’ public actions. In particular, it does not contain dependencies between public actions of other agents caused by their private facts. We propose a projection in which these private dependencies are maintained. The benefit of our dependency-preserving projection is demonstrated by using it to produce high-level plans in a new privacy-preserving planner, and as a heuristic for guiding forward search privacy-preserving algorithms. Both are able to solve more benchmark problems than any other state-of-the-art privacy-preserving planner. This more informed projection does not explicitly expose any private fact, action, or precondition. In addition, we show that even if an adversary agent knows that an agent has some private objects of a given type (e.g., trucks), it cannot infer the number of such private objects that the agent controls. This introduces a novel form of strong privacy, which we call object-cardinality privacy, that is motivated by real-world requirements.     

Specialized fast algorithms for IQC feasibility and optimization problems

The conventional way to treat integral quadratic constraint (IQC) problems is to transform them into semi-definite programs (SDPs). SDPs can then be solved using interior point methods which have been proven efficient. This approach, however, is not always the most efficient since it introduces additional decision variables to the SDP, and the additional decision variables sometimes largely increase the complexity of the problem. In this paper, we demonstrate how to solve IQC problems by other alternatives. More specifically, we consider two cutting plane algorithms. We will show that in certain cases these cutting plane algorithms can solve IQC problems much faster than the conventional approach. Numerical examples, as well as some explanations from the point of view of computational complexity, are provided to support our point.     

Parameter tuning of PBIL and CHC evolutionary algorithms applied to solve the Root Identification Problem

Evolutionary algorithms are among the most successful approaches for solving a number of problems where systematic searches in huge domains must be performed. One problem of practical interest that falls into this category is known as The Root Identification Problem in Geometric Constraint Solving, where one solution to the geometric problem must be selected among a number of possible solutions bounded by an exponential number. In previous works we have shown that applying genetic algorithms, a category of evolutionary algorithms, to solve the Root Identification Problem is both feasible and effective.In this work, we report on an empirical statistical study conducted to establish the influence of the driving parameters in the PBIL and CHC evolutionary algorithms when they are used to solve the Root Identification Problem. We identify a set of values that optimize algorithms performance. The driving parameters considered for the PBIL algorithm are population size, mutation probability, mutation shift and learning rate. For the CHC algorithm we studied population size, divergence rate, differential threshold and the set of best individuals. In both cases we applied unifactorial and multifactorial analysis, post hoc tests and best parameter level selection. Experimental results show that CHC outperforms PBIL when applied to solve the Root Identification Problem.     

Decision on the memory size of embedded information systems in an ubiquitous maintenance environment

Over the last decade, emerging information communication technologies have changed our stereotype of manufacturing and service companies. Now products equipped with embedded systems can be wirelessly networked, which leads to gathering and analyzing product status, and taking appropriate actions for maintenance operations during product lifecycle in an ubiquitous way. In this environment, it is necessary to determine the appropriate memory size of embedded systems for minimizing total maintenance system costs because the memory cost is a main cost factor for implementing the ubiquitous maintenance environment. We call it memory size decision problem in this study. We have formulated this problem with a non-linear model having constraints. The decision variable is the memory size of each embedded system. To solve this problem, we have proposed a meta heuristic search method based on genetic algorithms. To show the usefulness of the proposed heuristic, we have carried out computational experiments.     

Domain-independent feature extraction for multi-classification using multi-objective genetic programming

We propose three model-free feature extraction approaches for solving the multiple class classification problem; we use multi-objective genetic programming (MOGP) to derive (near-)optimal feature extraction stages as a precursor to classification with a simple and fast-to-train classifier. Statistically-founded comparisons are made between our three proposed approaches and seven conventional classifiers over seven datasets from the UCI Machine Learning database. We also make comparisons with other reported evolutionary computation techniques. On almost all the benchmark datasets, the MOGP approaches give better or identical performance to the best of the conventional methods. Of our proposed MOGP-based algorithms, we conclude that hierarchical feature extraction performs best on multi-classification problems.     

Novel algorithms of particle swarm optimisation with decision criteria

The particle swarm optimisation (PSO) is a stochastic, optimisation technique based on the movement and intelligence of swarms. In this paper, three new effective optimisation algorithms BPSO, HPSO and WPSO, by incorporating some decision criteria into PSO, have been proposed and analysed both in terms of their efficiency, resistance to the problem of premature convergence and the ability to avoid local optima. In the new algorithms, for each particle except position, two sets of velocities are generated and the profit matrix is constructed. Using the decision criteria the best strategy is selected. Simulations for benchmark test nonlinear function show that the algorithms in which the decision criteria have been applied, are beneficial over classical PSO in terms of their performance and efficiency.     

Performance evaluation of evolutionary heuristics in dynamic environments

In recent years, there has been a growing interest in applying genetic algorithms to dynamic optimization problems. In this study, we present an extensive performance evaluation and comparison of 13 leading evolutionary algorithms with different characteristics on a common platform by using the moving peaks benchmark and by varying a set of problem parameters including shift length, change frequency, correlation value and number of peaks in the landscape. In order to compare solution quality or the efficiency of algorithms, the results are reported in terms of both offline error metric and dissimilarity factor, our novel comparison metric presented in this paper, which is based on signal similarity. Computational effort of each algorithm is reported in terms of average number of fitness evaluations and the average execution time. Our experimental evaluation indicates that the hybrid methods outperform the related work with respect to quality of solutions for various parameters of the given benchmark problem. Specifically, hybrid methods provide up to 24% improvement with respect to offline error and up to 30% improvement with respect to dissimilarity factor by requiring more computational effort than other methods.