Tirtharaj Dash

An effective graph-theoretic approach towards simultaneous detection of fault(s) and cut(s) in wireless sensor networks

Summary A failed sensor node partitions a wireless sensor network (WSN) into 2 or more disjoint components, which is called as a cut in the network. The cut detection has been considered as a very challenging problem in the WSN research. In this paper, we propose a graph-theoretic distributed protocol to detect simultaneously the faults and cuts in the WSN. The proposed approach could be accomplished mainly in 3 phases, such that initialization phase, fault detection phase, and a cut detection phase. The protocol is an iterative method where at every time iteration, the node updates its state to calculate the potential factor. We introduced 2 terminologies such as a safe zone or cut zone of the network. The proposed method has been evaluated regarding various performance evaluation measures by implementing the same in the network simulator NS–2.35. The obtained results show that the proposed graph-theoretic approach is simple yet very powerful for the intended tasks.

Controlling Wall Following Robot Navigation Based on Gravitational Search and Feed Forward Neural Network

Automated control of mobile robot navigation is a challenging area in the field of robotics research. In this work, an attempt is made to use a new neural network training algorithm based on gravitational search (GS) and feed forward neural network (FFNN) for automatic robot navigation of wall following mobile robots. The GS strategy is used for setting the optimal weight set of the FFNN so as to increase the performance of the neural network. The algorithm is tested with three large datasets obtained from UCI machine learning repository, containing a sequence of sensor readings where sensors are arranged around the waist of the SCITOS G5 robot. The proposed method shows promising results for all the datasets.

Hybrid Gravitational Search and Particle Swarm Based Fuzzy MLP for Medical Data Classification

In this work, a hybrid training algorithm for fuzzy MLP, called Fuzzy MLP-GSPSO, has been proposed by combining two meta-heuristics: gravitational search (GS) and particle swarm optimization (PSO). The result model has been applied for classification of medical data. Five medical datasets from UCI machine learning repository are used as benchmark datasets for evaluating the performance of the proposed ‘Fuzzy MLP-GSPSO’ model. The experimental results show that Fuzzy MLP-GSPSO model outperforms Fuzzy MLP-GS and Fuzzy MLP-PSO for all the five datasets in terms of classification accuracy, and therefore can reduce overheads in medical diagnosis.

Gradient gravitational search: An efficient metaheuristic algorithm for global optimization

The adaptation of novel techniques developed in the field of computational chemistry to solve the concerned problems for large and flexible molecules is taking the center stage with regard to efficient algorithm, computational cost and accuracy. In this article, the gradient‐based gravitational search (GGS) algorithm, using analytical gradients for a fast minimization to the next local minimum has been reported. Its efficiency as metaheuristic approach has also been compared with Gradient Tabu Search and others like: Gravitational Search, Cuckoo Search, and Back Tracking Search algorithms for global optimization. Moreover, the GGS approach has also been applied to computational chemistry problems for finding the minimal value potential energy of two‐dimensional and three‐dimensional off‐lattice protein models. The simulation results reveal the relative stability and physical accuracy of protein models with efficient computational cost.

GA Based Polynomial Neural Network for Data Classification

Polynomial Neural Network is a self-organizing network whose performance depends strongly on the number of input variables and the order of polynomial which are determined by trial and error. In this paper, a training algorithm for Polynomial Neural Network (PNN) based on Genetic Algorithm (GA) has been proposed for classification problems. A performance comparison of the proposed PNN-GA and Back Propagation based PNN (PNN-BP) has also been carried out by considering four popular datasets obtained from UCI machine learning repository. Experimental results show that the proposed PNN-GA outperforms PNN-BP for all the four datasets and thus may be applied as classification model in many real world problems.

Neural network based automated detection of link failures in wireless sensor networks and extension to a study on the detection of disjoint nodes

In the broad research area of wireless sensor networks (WSN), detection of link failure is still in its infancy. In this paper, we propose to use a neural network model for detection of link failure in WSN. The neural network has been allowed to learn and adapt with the help of gradient descent based learning algorithm. We demonstrate the proposed model with regard to the preparation of training data and implementation of the model. This paper also provides a thorough theoretical and analytical investigation of link failures in WSN. The proposed neural network based model has been evaluated carefully with regard to testbed experiments. The simulation-based experiment has been conducted to justify the applicability of the proposed model for dense networks that could contain around 1000 links. We also analyze the theoretical performance of the proposed neural network based algorithm with regard to various performance evaluative measures such as failure detection accuracy, false alarm rate. The simulated experiments, as well as the testbed experiments in indoor and outdoor environments, suggest that the method is capable of link failure detection with higher detection rate and it is consistent. Furthermore, this article also reports a comprehensive case study as an extension of this present research towards automated detection of disjoint and disconnected nodes in a sensor network.

Investigation of RBF Kernelized ANFIS for Fault Diagnosis in Wireless Sensor Networks

Wireless sensor networks (WSN) are often deployed in human inaccessible environments. Some examples of such environments that are difficult for quick human reach include deep forests, various hazardous industries, hilltops, and sometimes underwater. The occurrence of failures in sensor networks is inevitable due to continuous or instant change in environmental parameters. A failure may lead to faulty readings which in turn may cause economic and physical damages to the environment. In this work, a thorough investigation has been conducted on the application of adaptive neuro-fuzzy inference system (ANFIS) for automated fault diagnosis in WSN. Further, a kernelized version of ANFIS has also been studied for the discussed problem. To avoid the model’s undesired biases toward a specific type of failure, oversampling has been done for multiple version of the ANFIS model. This study would serve as a guideline for the community toward the application of fuzzy inference approaches for fault diagnosis in sensor networks. However, the work focuses on the automated fault diagnosis in open air WSN and has no applicability in underwater sensor network systems.

Evolutionary Neural Networks versus Adaptive Resonance Theory Net for Breast Cancer Diagnosis

Breast cancer is the most common cancer among human females worldwide. Early detection of breast cancer is the only solution to reduce the breast cancer mortality. Machine learning models such as neural networks are one of the well-studied tools for the early detection of breast cancer. In this work, we implemented a supervised model called MLP and an unsupervised model called ART-1 net. We trained the MLP with two different evolutionary optimization techniques such as Biogeography-based Optimization (BBO) and Particle Swarm Optimization (PSO). The resulting models are called BBOMLP and PSOMLP. We compared the performance of these three models with Wisconsin Breast Cancer dataset. Our simulation results show that the unsupervised model could achieve better performance than the implemented supervised models.

Chain Multiplication of Dense Matrices: Proposing a Shared Memory based Parallel Algorithm

Chain multiplication of matrices is widely used for scientific computing. It becomes more challenging when there is large number of floating point dense matrices. Because, floating point operations take more time than integer operations. It would be interesting to lower the time of such chain operations. Now-a-days every multicore processor system has built in parallel computational power. This power can only be utilized when compatible parallel algorithms were used. So, in this work, a shared memory based parallel algorithms has been proposed to compute the multiplication of a long sequence of dense matrices. The algorithms have been tested with long sequence of matrices as input. The approach has been with 2×10 flops. The input matrix sequence length was typically varied from 2 to 30. Maximum number of processors used was eight (Eight core processor). Different parameters like speedup, efficiency etc. were also noted. It was concluded that the parallel algorithms could achieve approximately 90% efficiency at best case. The algorithms also showed improved scalability.

Large-Scale Assessment of Deep Relational Machines

Deep Relational Machines (or DRMs) present a simple way for incorporating complex domain knowledge into deep networks. In a DRM this knowledge is introduced through relational features: in the original formulation of [1], the features are selected by an ILP engine using domain knowledge encoded as logic programs. More recently, in [2], DRMs appear to achieve good performance without the need of feature-selection by an ILP engine (the features are simply drawn randomly from a space of relevant features). The reports so far on DRMs though have been deficient on three counts: (a) They have been tested on very small amounts of data (7 datasets, not all independent, altogether with few 1000s of instances); (b) The background knowledge involved has been modest, involving few 10s of predicates; and (c) Performance assessment has been only on classification tasks. In this paper we rectify each of these shortcomings by testing on datasets from the biochemical domain involving 100s of 1000s of instances; industrial-strength background predicates involving multiple hierarchies of complex definitions; and on classification and regression tasks. Our results provide substantially reliable evidence of the predictive capabilities of DRMs; along with a significant improvement in predictive performance with the incorporation of domain knowledge. We propose the new datasets and results as updated benchmarks for comparative studies in neural-symbolic modelling.