Search

Despite many decades of research on high-performance Deep Neural Network (DNN) accelerators, their massive computational demand still requires resource-efficient, optimized and parallel architecture for computational acceleration. Contemporary hardware implementations of DNNs face the burden of excess area requirement due to resource-intensive elements such as multipliers and non-linear Activation Functions (AFs). This paper proposes DNN with reused hardware-costly AF by multiplexing data using shift-register. The on-chip quantized log 2 based memory addressing with an optimized technique is used to access input features, weights, and biases. This way the external memory bandwidth requirement is reduced and dynamically adjusted for DNNs. Further, high-throughput and resource-efficient memory elements for sigmoid activation function are extracted using the Taylor series and its order expansion have been tuned for better test accuracy. The performance is validated and compared with previous work for the MNIST dataset. Besides, the digital design of AF is synthesized at 45 nm technology node and physical parameters are compared with previous work. The proposed hardware reused architecture is verified for neural network 16:16:10:4 using 8-bit dynamic fixed-point arithmetic and implemented on Xilinx Zynq xc7z010clg400 SoC using 100 MHz clock. The implemented architecture uses 25% less hardware resources and consumes 12% less power without performance loss, compared to other state-of-the-art implementations, as lower hardware resources and power consumption are especially important for increasingly important edge computing solutions.

The availability of commercial wearable bio-sensors provides an opportunity for developing smart phone applications for real-time diagnosis that can be used to improve the health of the user. We propose a multi-level information fusion approach for learning a predictive model for blood pressure (BP) using electrocardiogram (ECG) sensor data. The approach fuses the information on five different levels: i) data collection, where data from multiple ECG sensors is collected; ii) feature extraction, where features are extracted from the collected data by different preprocessing methods; iii) information fusion, fusing the evaluation information from different classifiers; iv) information fusion using the information from multi-target regression models for each BP class; and v) information fusion using the information from multi-target regression models from all configurations as a single model. This is used for predicting the blood pressure values (systolic BP (SBP), diastolic BP (DBP), and mean arterial pressure (MAP)). Evaluating the methodology by using a separate test set indicates that the multi-level information fusion provides promising results, which are acceptable and comparable to the state-of-the-art results obtained for blood pressure prediction.

This paper presents the perception of the quality of life and its relation to the digitalisation process in healthcare and the health tourism system. Health tourism system is part of the business transformation process inducted by the new technological improvements. Through this paper, we wont to understand better how the process of digitalisation affects the quality of life. This paper deals with the influence and applicability of sensor measurements in the creation of a new offer for innovative tourism and its reflection on tourism quality. It discusses the potential of measuring the impact of environmental colors on human body conditions to achieve the best solutions for wellbeing.

In memetic algorithm, a population based global search technique is used to broadly locate good areas of the search space, while repeated usage of a local search heuristic is employed to locate optimum. Intuitively, evolutionary operators that generate individuals with genetic material inherited from the parents and improved performance ability should be the right option for improved performance of the algorithm in terms of time and solution quality. Evolutionary operators with such properties were devised and used in memetic algorithm for solving multi-objective matrix tri-factorization problem. It was shown, by comparing deterministic naive approach with two variants of memetic algorithm with different level of inheritance, that evolutionary operators do not improve performance in this case. Further analysis showed that even though proposed evolutionary operators inherit high fitness from its parents, local search does not perform well on such offspring which results in poor performance.

A strong necessity for faster mass transport with high capacities and frequent runs has put pressure on railway infrastructure. These changes require an improvement in the maintenance activities that need to be planned carefully to reduce as much as possible their impacts on the actual use of the infrastructure. As a consequence, it is crucial to continuously monitor the technical equipment of the railroad tracks in the most efficient way. Railway switches operate in harsh environmental conditions. However, their reliability requirements are high due to safety and economic factors. Their maintenance depends on the data collected and the decisions on due corrective actions. The more frequent the data collection, the shorter the decision cycle, and the more intuitive the data visualization, the better the confidence in the system. This paper presents some tangible results, where the MANTIS concepts were applied to the use-case of railway switches to introduce continuous monitoring and proactive maintenance of the railroad tracks. The presented results include the measurement method, the analysis of the measured data, the profiling and modelling of the physical behaviour of the switch, including the recognized need to have separate models per season, and the visualization of the gathered data.

Studies of analytics integration management have largely focused on executive-led analytics at the enterprise level. However, in most organizations analytics initiatives do not enjoy executive support at the outset. Top management must first be convinced of the benefits, which slows down the path to competing via analytics. To successfully win top management support for broader analytics implementation an analytics pioneer should achieve five key aims: patiently build trust, manage interdisciplinary collaboration, focus on the problem solving action, facilitate the process, and importantly provide strong support to the embryonic analytics initiative. This is demonstrated through multiple-case study, presented in this paper. In embryonic analytics initiatives, the analytics champion appears locally, at mid-management level, and is up against the complex task of overcoming the resistance of an established organization, with its existing people, processes, data, technology, and culture. We examined what could be learned about the management of people-related issues in embryonic analytics processes. We studied the approaches used and lessons learned by all significant groups of stakeholders with the aim of helping managers show the value of analytics to their executives and colleagues.

There are many commercially available sensors for acquiring electrocardiogram (ECG) signals, and the predictive analyses are extremely welcome for real-time monitoring in public healthcare. One crucial task of such analysis is the selection of the pre-processing parameters for the ECG signals that carry the most valuable information. For this reason, we extended our previous work by proposing a framework, known as ECGpp, which can be used for selecting the best pre-processing parameters for ECG signals (like signal length and cut-off frequency) that will be involved in predictive analysis. The novelty of the framework is in the evaluation methodology that is used, where an ensemble of performance measures are used to rank and select the most promising parameters. Thirty different combinations of a signal length and a cut-off frequency were evaluated using a data set that contains data from five commercially available ECG sensors in the case of blood pressure classification. Evaluation results show that a signal length of 30 s carries the most valuable information, while it was demonstrated that lower cut-off frequencies, where the ECG components overlap with the baseline wander noise, can also provide promising results. Additionally, the results were tested according to the selection of the performance measures, and it was shown that they are robust to inclusion and exclusion of correlated measures.

Sustainable development requires measures to decrease energy consumption and to avoid any harmful impact on the environment. In recent years, a lot of research has been focused on designing energy systems by emulating the natural ones. This can be regarded as a novel way to support the transformation to sustainable energy systems. The initial step in forming energy systems of thus type, by emulating the ecosystem approach, is the development of a model for rendering the current situation. To contribute to the solution of this problem, an ecosystem model combining the analysis, optimisation and simulation modules has been developed as part of this study. The commonly used models for energy systems are good predictors of the general energy dynamics and structures. However, they can be inadequate when it comes to describing or predicting the complex phenomena within energy systems, such as the large number of parameters, the many end-users, the potential of the technologies, the assessment of the various environmental impacts and the variety of resources. With the implementation of the ecosystem model there is further expanding of the knowledge of an advanced, self-organising methodology integrated within the model’s operation by emulating the natural system dynamics. The ecosystem model could become a valuable tool for developing sustainable energy systems, and allowing the development of the most suitable and sustainable energy system based on the local availability of energy sources.

Forecasting the residential natural gas demand for large groups of buildings is extremely important for efficient logistics in the energy sector. In this paper different forecast models for residential natural gas demand of an urban area were implemented and compared. The models forecast gas demand with hourly resolution up to 60 h into the future. The model forecasts are based on past temperatures, forecasted temperatures and time variables, which include markers for holidays and other occasional events. The models were trained and tested on gas-consumption data gathered in the city of Ljubljana, Slovenia. Machine-learning models were considered, such as linear regression, kernel machine and artificial neural network. Additionally, empirical models were developed based on data analysis. Two most accurate models were found to be recurrent neural network and linear regression model. In realistic setting such trained models can be used in conjunction with a weather-forecasting service to generate forecasts for future gas demand.

Because a real-time monitoring using electrocardiogram (ECG) signals is a challenging task, the pre-processing techniques used for ECG signal analysis are crucial for obtaining information that is further used for some more complex analysis, such as predictive analyses. We compared different settings of parameters needed for pre-processing of ECG signals in order to estimate the valuable information that can be further used for blood pressure classification. Two parameters were involved in the comparison: i) the signal length used for ECG segmentation; and ii) the cut-off frequency used for baseline removal. The first parameter is the parameter used for obtaining ECG segments that are further used, and the second one is the frequency used for baseline removal. Thirty different combinations, each a combination of a signal length and a cut-off frequency, were evaluated using a dataset that contains data from five commercially available ECG sensors. For signal lengths: 10 s, 20 s, and 30 s, were used for data segmentation, while the cut-off frequency for baseline removal starts from 0.05 Hz, till 0.50 Hz, with a step length of 0.05 Hz. The evaluation of these combinations was done in combination with complexity analysis used for features extraction that are further used for blood pressure classification. Experimental results, obtained using a data-driven approach by comparing the combinations using the results obtained from the classification for 17 performance measures, showed that a signal length of 30 s carries the most information in a combination with cut-off frequency between 0.10 Hz and 0.20 Hz. Results contribute to the arguments published in the literature discussing the optimal ECG sample lengths needed for building predictive models, as well as the lower frequencies where the ECG components overlap with the baseline wander noise.

Deep neural networks are constructed that are able to partially solve a protein structure optimization problem. The networks are trained using reinforcement learning approach so that free energy of predicted protein structure is minimized. Free energy of a protein structure is calculated using generalized three-dimensional AB off-lattice protein model. This methodology can be applied to other classes of optimization problems and represents a step toward automatic heuristic construction using deep neural networks. Trained networks can be used to construct better initial populations for optimization. It is shown that differential evolution applied to protein structure optimization problem converges to better solutions when initial population is constructed in this way.

One of the most important components for proper functioning of cities is the mobility. As the paper focuses on transportation, the use of Intelligent Transportation Systems is crucial to enable modern smart cities solutions, both in cargo as well as passenger transport. This paper presents various problems of the typical city with regards to transport and mobility. Some problems are already partly tackled, but there are still open issues for research and development of even better approaches. Europe's Smart Specialization Strategy suggests that each region specializes in specific solutions, based on existing competences. The presented solutions are part of the Slovenian smart specialization strategy. However, the most important issue for implementation and for improving the real efficiency of the transportation in the smart city is simply the critical mass/acceptance rate of users/citizens.

Prediction of natural gas consumption of a town is very important for its logistics and district heating workflow. In this paper different empirical models are introduced for gas consumption prediction. The models are based on the characteristics of gas consumption data that was gathered in Ljubljana, Slovenia. Important observables were recognized from the data and the models were structured according to the correlation found in the data. The models are tested on forecasted weather data up to 60 hours into the future so that they can be used in realistic setting. The models presented in this paper are also able to model rare events such as public and school holidays.

Within highly competitive business environments, data mining (DM) is viewed as a significant technology to enhance decision-making processes by transforming data into valuable and actionable information to gain competitive advantage. There appears, however, to be a dearth of empirical case studies which consider in detail the initial stages in DM management to enable apt foundation for its later successful implementation. Our research applied a multi-method strategy to determine the critical success factors of embryonic DM implementation. We propose and validate, through a series of cases, a conceptual framework to guide practitioners' adoption of DM. Our findings reveal additional issues for applied decision making in the context of DM success.

The design optimization of an axial flux permanent magnet synchronous machine with a coreless stator and double external rotor is accomplished by using evolutionary optimization with a genetic algorithm and an analytical evaluation of objective functions. On the basis of eight variable geometry parameters, five objective functions are optimized in order to determine the maximum volume torque density and weight torque density, the minimum volume and weight of permanent magnets per Newton–meter, and the minimum machine price per Newton–meter. Based on the geometric parameters for minimum machine price per Newton–meter, a prototype is built for the rated torque. Optimized and analytically evaluated machine characteristics are validated with a finite-element method (FEM) and the measurements of a prototype. Evolutionary optimization with the analytical evaluation of objective functions significantly shortens the computational time required for design optimization in comparison with the FEM.

Planning problems can be solved with a large variety of different approaches, and a significant amount of work has been devoted to the automation of planning processes using different kinds of methods. This paper focuses on the use of specific local search algorithms for real-world production planning based on experiments with real-world data, and presents an adapted local search guided by evolutionary metaheuristics. To make algorithms efficient, many specifics need to be considered and included in the problem solving. We demonstrate that the use of specialized local searches can significantly improve the convergence and efficiency of the algorithm. The paper also includes an experimental study of the efficiency of two memetic algorithms, and presents a real-world software implementation for the production planning.

This paper presents a design approach for improvement of the efficiency of a universal motor, the type of motor that is typically used in home appliances and power tools. The goal of our optimization was to find optimal values of the independent geometrical parameters of the rotor and the stator of the UM with the aim of reducing the motor's main power losses-they occur in the iron and the copper. Our procedure is based on a genetic algorithm (GA), and by using it we were able to significantly improve the motor's efficiency. The GA proved to be a simple and efficient search-and-optimization method for solving this day-to-day design problem in industry. It significantly outperformed a conventional design procedure that was used previously.

We present a new design procedure that improves the efficiency of a universal motor, the type of motor that is typically used in home appliances and power tools. The goal of our optimization was to find the independent geometrical parameters of the rotor and the stator with the aim of reducing the motor's power losses, which occur in the iron and the copper. Our procedure is based on a genetic algorithm (GA) and by using this procedure we were able to significantly improve the motor's efficiency-the ratio of the motor's output power to its input power. The GA proved to be a simple and efficient search-and-optimization method for solving this day-to-day design problem in industry. It significantly outperformed a conventional "direct" design procedure that we had used previously.

This paper compares two metaheuristic approaches in solving a constrained transportation scheduling problem, which can be found in transporting goods in emergency situations. We compared Greedy Search, Parameter-less Evolutionary Search and Ant-Stigmergy Algorithm. The transportation scheduling/allocation problem is NP-hard, and is applicable to different real-life situations with high frequency of loading and unloading operations; like in depots, warehouses and ports. To evaluate the efficiency of the presented approaches, they were tested with four tasks based on realistic data. Each task was evaluated using group and free transportation approach. The experiments proved that all tested algorithms are viable option in solving such scheduling problems, however some performing better than others on some tasks. First Published Online: 03 Apr 2013

An initiative was introduced in one of the production facilities of Germany's E.G.O. Group in order to enhance its SAP information system with a custom-made application for production-scheduling optimization. The goal of the optimization is to find a production schedule that satisfies different, contradictory production and business constraints. We show the challenges faced in the application of the multi-objective optimization approach, which is gaining influence in the management of production scheduling. We implement a memetic version of the Indicator-Based Evolutionary Algorithm with customized reproduction operators and local search procedures to find a set of feasible, non-dominated solutions. Such a memetic algorithm was applied to two real order lists from the production company. Additionally, we also lay out an efficient presentation of the multi-objective results for an expert's support in decision making. This provides the management with the possibility to gain additional insights into how the production schedule dynamically reacts to changes in the decision criteria. We show that the multi-objective approach is able to find high-quality solutions, which enables flexibility when it comes to quickly adapting to specific business conditions.

In this paper we present a program tool that was developed for the simulation of temperatures inside the cooling appliance. The simulator is used to check the performance of the appliance at different settings of control parameters. These simulations replace the part of the measurements, which are time consuming due to the slow thermal processes, but are needed in the appliance development process. The simulator was designed in Microsoft Excel. The accuracy validation of the simulation shows the reliability of the developed simulator.

The aim of the paper is to present a critical review of analytics and visualisation technology for big data, and propose future directions to overcome the shortcomings of the current technologies. The current machine learning and data-mining algorithms are operating mostly on predefined scales of aggregation, while in the vast amounts of data the problem arises at the level of aggregation which cannot be defined ahead of time. We therefore identify a novel and extended architecture to operate on flexible multi-resolution hypothesis space. With such architecture framework the goal is to open a space of possibly discovered models towards classes of data, which are by today's approaches discovered only for special cases. Furthermore, the multi-resolution approach to big-data analytics could allow scenarios like semi-supervised and unsupervised anomaly detection, detecting complex relationships from the heterogeneous data sources, and providing ground for visualisation of complex processes.

The main motivation for organizations to e-collaborate is to enable knowledge sharing and learning in order to effectively address a new business opportunity by forming a Virtual Organization (VO) for solving the given task. One of the difficulties in VO creation is appropriate partner selection with mutual trust, as well as the support for the management of trust in a broader Virtual organization Breeding Environment (VBE)- a cluster of organizations willing to collaborate when a new business opportunity appears. This paper proposes an approach to modeling trust in a network of collaborating organizations, aimed at improved trust management in VBEs and improved decision support in the process of VO creation.

This paper presents a new procedure for designing the rotor and the stator of a universal motor for home appliances that is based on a genetic algorithm (GA). The GA was employed in order to optimize the independent geometric parameters of the rotor/stator lamination with the aim to reduce the main motor's power losses, i.e. the losses occurring in the iron and the copper. Using this procedure the motor's technical quality, expressed as efficiency (i.e. the ratio of the motor's output power to its input power), can be significantly improved. The GA proved to be a simple and efficient search-and-optimization method for solving this day-to-day design problem in industry. It significantly outperformed a conventional 'direct' design procedure that had been used before.

In this paper, we present a methodology for learning elementary-school mathematics online. It is supported by a decision making system, based on evolutionary computation, which leads a student in selecting an optimal subset of math items to effectively upgrade the knowledge. We evaluated the system's effectiveness with the active participation of experienced teachers and students. The teachers evaluated solutions provided by the systems, judging the selection with respect to content areas and difficulty levels. The students were solving math items suggested by the system, gradually improving their knowledge. They both agreed that the system could effectively be used for e-learning.

When dealing with real-world problems, it turns out that there are many specifics of the problem we are trying to solve. Since many algorithms that are being developed are evaluated and compared on test benchmark problems, they can simulate real-world problems up to some degree and specifics are not presumed and tested. To make algorithms efficient, such specifics need to be considered and included in the problem solving. In this paper, a real-world production scheduling problem is addressed. A typical approach with genetic algorithm turned out to be insufficient due to added complexity of many specifics. To successfully solve this problem, a memetic algorithm, which uses problem-specific local search procedures to improve solutions acquired by genetic algorithm, is proposed. It is shown that the use of such local search procedures can significantly improve the effectiveness and efficiency of the algorithm.

Most digital circuits synthesized from a finite state machine contain idle functional units (modules) at some time during operation, in the sense that they do not contain any valid data. In a synchronous circuit with an on-chip synthesized controller which explicitly determines the flow of data, the idle functional units are known in advance for each FSM stage. A case study of the implementation of an online built-in self test that exploits the idle modules is presented. A concurrent test data path which cycles test vectors through the idle functional units and produces a response signature is synthesized along with the original algorithm, avoiding test time overheads.

This paper presents a new designing procedure for the rotor and the stator of a universal motor for home appliances. It is based on a genetic algorithm (GA). The aim is to optimize the independent geometric parameters of the rotor/stator lamination to reduce the main motor's power losses, i.e., the iron and the copper losses. The GA proved to be a simple and efficient search and optimization method for solving the daily design problem in the industry. Using this procedure the motor's technical quality, expressed as efficiency, has been significantly improved.

The concept of the MANTIS project is to provide a proactive maintenance service platform architecture based on Cyber Physical Systems that allows to estimate future performance, to predict and prevent imminent failures and to schedule proactive maintenance. Maintenance is an important function that creates additional value in the business process as well as new business models with a stronger service orientation. Physical systems and the environment they operate in are monitored continuously by a broad and diverse range of intelligent sensors, resulting in massive amounts of data that characterize the usage history, operational condition, location, movement and other physical properties of those systems. These systems form part of a larger network of heterogeneous and collaborative systems (e.g. vehicle fleets or photovoltaic parks) connected via robust communication mechanisms able to operate in challenging environments. MANTIS consists of distributed processing chains that efficiently transform raw data into knowledge while minimizing the need for bandwidth. Sophisticated distributed sensing and decision making functions are performed at different levels in a collaborative way, ranging from local nodes to locally optimize performance, bandwidth and maintenance; to cloud-based platforms that integrate information from diverse systems and execute distributed processing and analytics algorithms for global decision making. MANTIS will contribute to companies' assets availability, competitiveness, growth and sustainability.