Showing papers presented at "Computer Aided Systems Theory in 2017"

On Sharp Bounds on the Rate of Convergence for Finite Continuous-Time Markovian Queueing Models

Abstract: Finite inhomogeneous continuous-time Markov chains are studied. For a wide class of such processes an approach is proposed for obtaining sharp bounds on the rate of convergence to the limiting characteristics. Queueing examples are considered.

Overtaking Maneuver for Automated Driving Using Virtual Environments

Abstract: Among the driving possible scenarios in highways, the overtaking maneuver is one of the most challenging Its high complexity along with the interest in automated cooperative vehicles make this maneuver one of the most studied topics on the field on last years It involves a great interaction between both longitudinal (throttle and brake) and lateral (steering) actuators This work presents a three phases overtaking path planning using Bezier curves, with special interest in the continuity of the curvature Communication among the vehicles is also considered Finally, the maneuver will be validated using Dynacar, a dynamic model vehicle simulator

Exploring Food Detection Using CNNs

Abstract: One of the most common critical factors directly related to the cause of a chronic disease is unhealthy diet consumption. Building an automatic system for food analysis could enable a better understanding of the nutritional information associated to the food consumed and thus, help taking corrective actions on our diet. The Computer Vision community has focused its efforts on several areas involved in visual food analysis such as: food detection, food recognition, food localization, portion estimation, among others. For food detection, the best results in the state of the art were obtained using Convolutional Neural Networks. However, the results of all different approaches were tested on different datasets and, therefore, are not directly comparable. This article proposes an overview of the last advances on food detection and an optimal model based on the GoogLeNet architecture, Principal Component Analysis, and a Support Vector Machine that outperforms the state of the art on two public food/non-food datasets.

On Fractional Stochastic Modeling of Neuronal Activity Including Memory Effects

Abstract: In order to model the memory and to describe the memory effects in the firing activity of a single neuron subject to a time-dependent input current, a fractional stochastic Langevin-type equation is considered. Two different discretization formulas are derived and the corresponding algorithms are implemented by means of R-codes for several values of the parameters. Reset mechanisms after successive spike times are suitably imposed to compare simulation results. The firing rates and some neuronal statistical estimates obtained by means the two algorithms are provided and discussed.

Tabu Search and Solution Space Analyses. The Job Shop Case

Abstract: We present some own results of the theoretical and experimental investigations of the landscape of the solution space for the job-shop scheduling problem. Provided interpretation of the space throws new light on the process of solving hard combinatorial optimization problems as well as on Tabu Search phenomenon.

Towards Egocentric Sentiment Analysis

Abstract: The availability and use of egocentric data are rapidly increasing due to the growing use of wearable cameras. Our aim is to study the effect (positive, neutral or negative) of egocentric images or events on an observer. Given egocentric photostreams capturing the wearer’s days, we propose a method that aims to assign sentiment to events extracted from egocentric photostreams. Such moments can be candidates to retrieve according to their possibility of representing a positive experience for the camera’s wearer. The proposed approach obtained a classification accuracy of 75% on the test set, with deviation of 8%. Our model makes a step forward opening the door to sentiment recognition in egocentric photostreams.

Heartbeat Classification of ECG Signals Using Rational Function Systems

Abstract: The main idea of this paper is to show that rational orthogonal function systems, called Malmquist-Takenaka (MT) systems can effectively be used for ECG heartbeat classification The idea behind using these systems is the adaptive nature of them Then the constructed feature vector consists of two main parts, called dynamic and morphological parameters The latter ones contain the coefficients of the orthogonal projection with respect to the MT systems Then Support Vector Machine algorithm was used for classifying the heartbeats into the usual 16 arrhythmia classes The comparison test were performed on the MIT-BIH arrhythmia database The results show that our algorithm outperforms the previous ones in many respects

Optimization of Passenger Distribution at Metro Stations Through a Guidance System

Abstract: The steady growth of population in cities demands an efficient subway management system. To alleviate crowding on certain trains and subway lines, especially during rush hours, we propose a system which optimizes passenger distribution. Through visual cues directly displayed on the waiting platform, the passengers are informed about the wagons’ occupation rates, so that they can make a decision about which wagon to use before the subway arrives. Experimental results show that the relevance of the implementation is significant for both passenger satisfaction as well as train operators. In combination with a graphical user interface, the advantages of a guidance system could be demonstrated. The acceptance of our system was guaranteed by 75% of the passengers questioned, who stated they would use such a guidance system.

Mobile Wrist Vein Authentication Using SIFT Features

Abstract: Biometrics have become important for authentication on modern mobile devices. Thereby, different biometrics are differently hard to observe by attackers: for example, veins used in vein pattern authentication are only revealed with specialized hardware. In this paper we propose a low cost mobile vein authentication system based on Scale-Invariant Feature Transform (SIFT). We implement our approach as vein recording and authentication prototype, evaluate it using a self recorded vein database, and compare results to other vein recognition approaches applied on the same data.

Particle Therapy Patient Scheduling: Time Estimation for Scheduling Sets of Treatments

Abstract: In the particle therapy patient scheduling problem (PTPSP) cancer therapies consisting of sequences of treatments have to be planned within a planning horizon of several months. In our previous works we approached PTPSP by decomposing it into a day assignment part and a sequencing part. The decomposition makes the problem more manageable, however, both levels are dependent on a large degree. The aim of this work is to provide and a surrogate objective function that quickly predicts the behavior of the sequencing part with reasonable precision, allowing an improved day assignment w.r.t. the original problem.

Review of UHF-Based Signal Processing Approaches for Partial Discharge Detection

Abstract: Partial Discharge (PD) events are due to local defects in dielectrics and can cause damages to the electrical insulation and eventually to the whole power station This paper reviews approaches describing procedures and numerical techniques for detecting, denoising, clustering, and classifying PDs in the ultra-high frequency range For each method the mathematical background is recalled and one or few representative examples from selected papers are shortly described

Rational Variable Projection Methods in ECG Signal Processing

Abstract: In this paper we develop an adaptive electrocardiogram (ECG) model based on rational functions. We approximate the original signal by the partial sums of the corresponding Malmquist–Takenaka–Fourier series. Our aim in the construction of the model was twofold. Namely, besides good approximation an equally important point was to have direct connection with medical features. To this order, we consider the rational optimization problem as a special variable projection method. Based on the natural segmentation of a heartbeat into P, QRS, T waves, we use three complex parameters, i.e. the poles of the rational functions. For the optimization of the parameters, we apply constrained optimization. As a result every pole corresponds to one of these waves. We developed and tested our method by using the MIT-BIH Arrhythmia Database.

On a Non-homogeneous Gompertz-Type Diffusion Process: Inference and First Passage Time

Abstract: A stochastic diffusion model based on a generalized Gompertz deterministic growth in which the carrying capacity depends on the initial size of the population is considered. The growth parameter of the process is then modified by introducing a time-dependent exogenous term. The first passage time problem is considered and a two-step procedure to estimate the model is proposed. Simulation study is also provided for suitable choices of the exogenous term.

Deep Reinforcement Learning in Serious Games: Analysis and Design of Deep Neural Network Architectures

Abstract: Serious games present a noteworthy research area for artificial intelligence, where automated adaptation and reasonable NPC behaviour present essential challenges. Deep reinforcement learning has already been successfully applied to game-playing. We aim to expand and improve the application of deep learning methods in SGs through investigating their architectural properties and respective application scenarios. In this paper, we examine promising architectures and conduct first experiments concerning CNN design and analysis for game-playing. Although precise statements about the applicability of different architectures are not yet possible, our findings allow for concluding some general recommendations for the choice of DL architectures in different scenarios. Furthermore, we point out promising prospects for further research.

Detecting Hands in Egocentric Videos: Towards Action Recognition

Abstract: Recently, there has been a growing interest in analyzing human daily activities from data collected by wearable cameras. Since the hands are involved in a vast set of daily tasks, detecting hands in egocentric images is an important step towards the recognition of a variety of egocentric actions. However, besides extreme illumination changes in egocentric images, hand detection is not a trivial task because of the intrinsic large variability of hand appearance. We propose a hand detector that exploits skin modeling for fast hand proposal generation and Convolutional Neural Networks for hand recognition. We tested our method on UNIGE-HANDS dataset and we showed that the proposed approach achieves competitive hand detection results.

One Degree of Freedom Copter

Abstract: The paper presents laboratory model of unmanned aerial vehicle. The plant design, construction and building was a part of a project in the course on autonomous mechatronic systems. The goal was to make an exercise in mechatronics, resulting in plants which could be used in other project or classes, using the same hardware as common UAVs.

Measures for the Evaluation and Comparison of Graphical Model Structures

Abstract: Structure learning is the identification of the structure of graphical models based solely on observational data and is NP-hard. An important component of many structure learning algorithms are heuristics or bounds to reduce the size of the search space. We argue that variable relevance rankings that can be easily calculated for many standard regression models can be used to improve the efficiency of structure learning algorithms. In this contribution, we describe measures that can be used to evaluate the quality of variable relevance rankings, especially the well-known normalized discounted cumulative gain (NDCG). We evaluate and compare different regression methods using the proposed measures and a set of linear and non-linear benchmark problems.

Solving the Traveling Thief Problem Using Orchestration in Optimization Networks

Abstract: Optimization problems can sometimes be divided into multiple subproblems. Working on these subproblems instead of the actual master problem can have some advantages, e.g. if they are standard problems, it is possible to use already existing algorithms, whereas specialized algorithms would have to be implemented for the master problem. In this paper we approach the NP-hard Traveling Thief Problem by implementing different cooperative approaches using optimization networks. Orchestration is used to guide the algorithms that solve the respective subproblems. We conduct experiments on some instances of a larger benchmark set to compare the different network approaches to best known results, as well as a sophisticated, monolithic approach. Using optimization networks, we are able to find new best solutions for all of the selected problem instances.

Facilitating Evolutionary Algorithm Analysis with Persistent Data Structures

Abstract: Evolutionary algorithm analysis is often impeded by the large amounts of intermediate data that is usually discarded and has to be painstakingly reconstructed for real-world large-scale applications. In the recent past persistent data structures have been developed which offer extremely compact storage with acceptable runtime penalties. In this work two promising persistent data structures are explored in the context of evolutionary computation with the hope to open the door to simplified analysis of large-scale evolutionary algorithm runs.

A Multi-encoded Genetic Algorithm Approach to Scheduling Recurring Radiotherapy Treatment Activities with Alternative Resources, Optional Activities, and Time Window Constraints

Abstract: The radiotherapy patient scheduling problem deals with the assignment of recurring treatment appointments to patients diagnosed with cancer. The appointments must take place at least four times within five consecutive days at approximately the same time. Between daily appointments, optional (imaging) activities that require alternative resources, also must be scheduled. A pertinent goal therefore is minimizing both the idle time of the bottleneck resource (i.e., the particle beam used for the irradiation) and the potential risk of a delayed start. To address this problem, we propose a multi-encoded genetic algorithm. The chromosome contains the assignment of treatments to days for each patient, information on which optional activities to schedule, and the patient sequence for each day. To ensure feasibility during the evolutionary process, we present tailored crossover and mutation operators. We also compare a chronological solution decoding approach and an algorithm that fills idle times between already scheduled activities. The latter approach outperforms chronological scheduling on real-world-inspired problem instances. Furthermore, forcing some of the offspring to improve the parent’s fitness (i.e., offspring selection) within the genetic algorithm is beneficial for this problem setting.

A Real-Time Classification System for Upper Limb Prosthesis Control in MATLAB

Abstract: In this paper we present a MATLAB tool for processing both EMG and NIR sensor signals in real-time in order to provide a fully operational tool for clinical testing After a short training phase, the decision tree classifier produces output for actuating a Michelangelo hand by Ottobock Healthcare To validate the system design, it was tested with four probands performing wrist flexion, wrist extension and fist hand movement patterns After a training phase, features were extracted in real-time from either the EMG or NIR sensor data for classification with the model created during the training phase In this setup, NIR sensor data alone proved to be sufficient for distinguishing three hand movement patterns with two sensors The classification accuracy is equal or better to standard EMG data recorded from the same sensor pick-up area on the forearm

Meta-learning-based system for solving logistic optimization problems

Abstract: The Algorithm Selection Problem seeks to select the most suitable algorithm for a given problem For solving it, the algorithm selection systems have to face the so-called cold start It concerns the disadvantage that arises in those cases where the system involved in the selection of the algorithm has not enough information to give an appropriate recommendation Bearing that in mind, the main goal of this work is two-fold On the one hand, a novel meta-learning-based approach that allows selecting a suitable algorithm for solving a given logistic problem is proposed On the other hand, the proposed approach is enabled to work within cold start situations where a tree-structured hierarchy that enables to compare different metric dataset to identify a particular problem or variation is presented

On the Imputation of Missing Values in Univariate \(PM_{10}\) Time Series

Abstract: Missing data frequently happen in environmental research, usually due to faults in data acquisition, inadequate sampling or measurement error They make difficult to determine whether the limits set by the European Community on certain indicators of air quality are fulfilled or not Indeed, due to missing values, the number of exceedances per year of \(PM_{10}\), that is particulate matter 10 \(\upmu \)m or less in diameter, and other air quality indicators are often heavily underestimated, and no environmental policy is applied to protect citizen health

Analysis of Schema Frequencies in Genetic Programming

Abstract: Genetic Programming (GP) schemas are structural templates equivalent to hyperplanes in the search space Schema theories provide information about the properties of subsets of the population and the behavior of genetic operators In this paper we propose a practical methodology to identify relevant schemas and measure their frequency in the population We demonstrate our approach on an artificial symbolic regression benchmark where the parts of the formula are already known Experimental results reveal how solutions are assembled within GP and explain diversity loss in GP populations through the proliferation of repeated patterns

GRASP-VNS for a Periodic VRP with Time Windows to Deal with Milk Collection.

Abstract: This paper considers the planning of the collection of fresh milk from local farms with a fleet of refrigerated vehicles. The problem is formulated as a version of the Periodic Vehicle Routing Problem with Time Windows. The objective function is oriented to the quality of service by minimizing the service times to the customers within their time windows. We developed a hybrid metaheuristic that combines GRASP and VNS to find solutions. In order to help the hybrid GRASP-VNS find high-quality and feasible solutions, we consider infeasible solutions during the search using different penalty functions.

Cost-Efficient Traffic Sign Detection Relying on Smart Mobile Devices

Abstract: Neglect of the instructions of road traffic signs is one of the main contributing factors in road accidents. Smartphone traffic sign detection technology can offer significant information about the driving environment and increase driving comfort and traffic safety. It could also have interesting road inventory and maintenance applications. In this paper, we propose a driver assistance system for real-time detection of traffic signs on smartphone platforms using the OpenCV computer vision library. This technology uses the back camera of a smartphone to capture images of the driving environment and then uses advanced image processing functions to detect traffic signs. The field experiment on target traffic signs showed an 85% detection rate. The performance of the application may vary between devices with different processing power and camera quality.

Non-email Spam and Machine Learning-Based Anti-spam Filters: Trends and Some Remarks.

Abstract: Electronic spam, or unsolicited and undesired messages sent massively, is one of the threats that affects email and other media The high volume and ratio of email spam have generated enormous time and economic losses Due to this, many different email anti-spam defenses have been used This translated into more complex spams in order to surpass them Moreover, the spamming business moved to the less protected yet quite profitable non-email media because of the numerous potential targets that results from their extensive usage Since that moment, spams in these media have increased rapidly in quantity, sophistication and danger, especially in the most popular ones: Instant Messaging, SMS and social media Therefore, in this paper some of the characteristics and statistics of instant spam, mobile spam and social spam are exposed Then, an overview of anti-spam techniques developed during the last decade to fight these new spam trends is presented, focusing on hybrid and Machine Learning-based approaches We conclude with some possible future evolutionary steps of both non-email spams and anti-spams

Modifications of Model Free Control to FOTD Plants

Abstract: Model free control (MFC) represents one of possible alternatives to traditional approaches as PID control, disturbance observer based control (DOBC), internal model control (IMC), etc. As one of its central features one could mention use of finite-impulse-response (FIR) filters in input disturbance reconstruction. The paper deals with characterizing links of MFC to these approaches and discussing the terminology used and simple procedures for tuning of such controllers for the plants approximated by the first order time delayed models.

Integrating Exploratory Landscape Analysis into Metaheuristic Algorithms

Abstract: The no free lunch (NFL) theorem puts a limit to the range of problems a certain metaheuristic algorithm can be applied to successfully. For many methods these limits are unknown a priori and have to be discovered by experimentation. With the use of fitness landscape analysis (FLA) it is possible to obtain characteristic data and understand why methods perform better than others. In past research this data has been gathered mostly by a separate set of exploration algorithms. In this work it is studied how FLA methods can be integrated into the metaheuristic algorithm. We present a new exploratory method for obtaining landscape features that is based on path relinking (PR) and show that this characteristic information can be obtained faster than with traditional sampling methods. Path relinking is used in several metaheuristic which creates the possibility of integrating these features and enhance algorithms to output landscape analysis in addition to good solutions.

Offspring Selection Genetic Algorithm Revisited: Improvements in Efficiency by Early Stopping Criteria in the Evaluation of Unsuccessful Individuals

Abstract: This paper proposes some algorithmic extensions to the general concept of offspring selection which itself is an algorithmic extension of genetic algorithms and genetic programming. Offspring selection is characterized by the fact that many offspring solution candidates will not participate in the ongoing evolutionary process if they do not achieve the success criterion. The algorithmic enhancements proposed in this contribution aim to early estimate if a solution candidate will not be accepted based on partial solution evaluation. The qualitative characteristics of offspring selection are not affected by this means. The discussed variant of offspring selection is analyzed for several symbolic regression problems with offspring selection genetic programming. The achievable gains in terms of efficiency are remarkable especially for large data-sets.