α− dissipative filtering for Singular Markov jump system with time delays

This paper considers the [Formula: see text] dissipative filtering problem for a class of Singular Markov jump systems (SMJSs) with distributed time delays and discrete time delays. First, using Lyapunov’s stability theory and combining delay partitioning technique, integral partitioning technique, and free weight matrix method, the sufficient conditions for stochastic admissibility and [Formula: see text] dissipation of system are studied. Then, a filtering design method based on linear matrix inequalities (LMIs) is given to make the filtering error system stochastically admissible and [Formula: see text] dissipative. Finally, numerical simulations verify the effectiveness of the resulting method.

Mathematical Modelling of Twin -T Notch Filter using Floating Admittance Matrix

A novel article presents the RC-notch filter function using the floating admittance matrix approach. The main advantages of the approach underlined the easy implementation and effective computation. The proposed floating admittance matrix (FAM) method is unique, and the same can be used for all types of electronic circuits. This method takes advantage of the partitioning technique for a large network. The sum property of all the elements of any row or any column equal to zero provides the assurance to proceed further for analysis or re-observe the very first equation at the first instant itself. This saves time and energy. The FAM method presented here is so simple that anybody with slight knowledge of electronics but understating the matrix maneuvering can analyze any circuit to derive all types of transfer functions. The mathematical modelling using the FAM method allows the designer to adjust their design at any stage of analysis comfortably. These statements provide compelling reasons for the adoption of the proposed process and demonstrate its benefits.

Mathematical Modelling and Simulation of Band Pass Filters using the Floating Admittance Matrix Method

This article aims to develop a band pass filter's mathematical model using the Floating Admittance Matrix (FAM) method. The use of the conventional methods of analysis based KCL, KVL, Thevenin's, Norton's depends on the type of the particular circuit. The proposed mathematical modeling using the floating admittance matrix method is unique, and the same can be used for all types of circuits. This method uses the partitioning technique for large network. The sum property of all the elements of any row or any column equal to zero provides the assurance to proceed further for analysis or re-observe the very first equation. This saves time and energy. The FAM method presented here is so simple that anybody with slight knowledge of electronics but understating the matrix maneuvering, can analyze any circuit to derive all types of transfer functions. The mathematical modeling using the FAM method provides leverage to the designer to comfortably adjust their design at any stage of analysis. These statements provide compelling reasons for the adoption of the proposed process and demonstrate its benefits

Communication Failure Resilient Distributed Neural Network for Edge Devices

Recently, the necessity to run high-performance neural networks (NN) is increasing even in resource-constrained embedded systems such as wearable devices. However, due to the high computational and memory requirements of the NN applications, it is typically infeasible to execute them on a single device. Instead, it has been proposed to run a single NN application cooperatively on top of multiple devices, a so-called distributed neural network. In the distributed neural network, workloads of a single big NN application are distributed over multiple tiny devices. While the computation overhead could effectively be alleviated by this approach, the existing distributed NN techniques, such as MoDNN, still suffer from large traffics between the devices and vulnerability to communication failures. In order to get rid of such big communication overheads, a knowledge distillation based distributed NN, called Network of Neural Networks (NoNN), was proposed, which partitions the filters in the final convolutional layer of the original NN into multiple independent subsets and derives smaller NNs out of each subset. However, NoNN also has limitations in that the partitioning result may be unbalanced and it considerably compromises the correlation between filters in the original NN, which may result in an unacceptable accuracy degradation in case of communication failure. In this paper, in order to overcome these issues, we propose to enhance the partitioning strategy of NoNN in two aspects. First, we enhance the redundancy of the filters that are used to derive multiple smaller NNs by means of averaging to increase the immunity of the distributed NN to communication failure. Second, we propose a novel partitioning technique, modified from Eigenvector-based partitioning, to preserve the correlation between filters as much as possible while keeping the consistent number of filters distributed to each device. Throughout extensive experiments with the CIFAR-100 (Canadian Institute For Advanced Research-100) dataset, it has been observed that the proposed approach maintains high inference accuracy (over 70%, 1.53× improvement over the state-of-the-art approach), on average, even when a half of eight devices in a distributed NN fail to deliver their partial inference results.

AN EXPERIMENTAL ANALYSIS OF SPATIAL INDEXING ALGORITHMS FOR REAL TIME SAFETY CRITICAL MAP APPLICATION

This paper presents a study that compares the three space partitioning and spatial indexing techniques, KD Tree, Quad KD Tree, and PR Tree. KD Tree is a data structure proposed by Bentley (Bentley and Friedman, 1979) that aims to cluster objects according to their spatial location. Quad KD Tree is a data structure proposed by Berezcky (Bereczky et al., 2014) that aims to partition objects using heuristic methods. Unlike Bereczky’s partitioning technique, a new partitioning technique is presented based on dividing objects according to space-driven, in the context of this study. PR Tree is a data structure proposed by Arge (Arge et al., 2008) that is an asymptotically optimal R-Tree variant, enables data-driven segmentation. This study mainly aimed to search and render big spatial data in real-time safety-critical avionics navigation map application. Such a real-time system needs to efficiently reach the required records inside a specific boundary. Performing range query during the runtime (such as finding the closest neighbors) is extremely important in performance. The most crucial purpose of these data structures is to reduce the number of comparisons to solve the range searching problem. With this study, the algorithms’ data structures are created and indexed, and worst-case analyses are made to cover the whole area to measure the range search performance. Also, these techniques’ performance is benchmarked according to elapsed time and memory usage. As a result of these experimental studies, Quad KD Tree outperformed in range search analysis over the other techniques, especially when the data set is massive and consists of different geometry types.

Hardware software partitioning using directed acyclic data dependence graph with precedence

n this these we present a system partitioning technique that employs C/C++ as input specification language for hardware/software co-design. The proposed algorithm is able to explore a number of partitioning solutions as compared to other partitioning research. This benefit is obtained by processing data dependency and precedence dependency simultaneously in a new representation called Directed Acyclic Data dependency Graph with Precedence (DADGP). DADGP is an extension of Directed Acyclic Graph (DAG) structure frequently used in the past for partitioning. The DADGP based partitioning algorithm minimizes communication overhead, overall system execution time as well as system cost in terms of hardware area. The algorithm analyzes the DADGP and tries to expose parallelism between processing elements and repeated tasks. The benefits of exposing parallelism with minimum inter PE communication overhead are shown in the experimental results. However, such benefits come with increase in cost due to additional hardware units and their interconnections. DADGP-based partitioning technique is also employed to implement block matching and SOBEL edge detection techniques. Overall, the proposed system partitioning algorithm is fast and powerful enough to handle complicated and large system designs.

Hardware software partitioning using directed acyclic data dependence graph with precedence

n this these we present a system partitioning technique that employs C/C++ as input specification language for hardware/software co-design. The proposed algorithm is able to explore a number of partitioning solutions as compared to other partitioning research. This benefit is obtained by processing data dependency and precedence dependency simultaneously in a new representation called Directed Acyclic Data dependency Graph with Precedence (DADGP). DADGP is an extension of Directed Acyclic Graph (DAG) structure frequently used in the past for partitioning. The DADGP based partitioning algorithm minimizes communication overhead, overall system execution time as well as system cost in terms of hardware area. The algorithm analyzes the DADGP and tries to expose parallelism between processing elements and repeated tasks. The benefits of exposing parallelism with minimum inter PE communication overhead are shown in the experimental results. However, such benefits come with increase in cost due to additional hardware units and their interconnections. DADGP-based partitioning technique is also employed to implement block matching and SOBEL edge detection techniques. Overall, the proposed system partitioning algorithm is fast and powerful enough to handle complicated and large system designs.

Improving Indoor Localization System Using a Partitioning Technique Based on RSS and ToA

Designing a localization system for an indoor environment faces more challenges because of multipath and interference problems. In this field, the most important techniques used for such environment, are RSS and ToA which need to be improved especially from more interference because of the huge multipath problems. In this paper, a case study of a selected building is chosen in order to apply the proposed technique of this research. Such proposal is based on the PT ‎ of the area in the case study into MZ. Each zone is allocated special values for the parameters used to estimate the target positions. WI package is used to simulate the case study area and apply such proposal based on RSS and ToA. The results confirm that the estimated locations are close to the real locations by the average error of (2.8) meter and (0.192) meter for ToA corresponding one zone and four zones ‎respectively. ‎ In contrast, the results of our experiment show that the accuracy is improved from an average error of (2.4) meter and (0.217) meter for RSS corresponding one zone and four zones ‎respectively‎. Such results confirm that dividing the case study area into more zones leads to more accuracy.