Shakti Kinger

Bio: Shakti Kinger is an academic researcher from Massachusetts Institute of Technology. The author has contributed to research in topics: Fault management & Fault coverage. The author has an hindex of 2, co-authored 4 publications receiving 11 citations. Previous affiliations of Shakti Kinger include Maharashtra Institute of Technology.

Probe station placement algorithm for probe set reduction in network fault localization

Abstract: Fault diagnosis is a central aspect of network fault management. Since faults are unavoidable in communication systems, their quick detection and isolation is essential for the robustness, reliability, and accessibility of a system. Probing technique for fault localization involves placement of probe stations (Probe stations are specially instrumented nodes from where probes can be sent to monitor the network) which affects the diagnosis capability of the probes sent by the probe stations. Probe station locations affect probing efficiency, monitoring capability, and deployment cost. At the same time, probe station selection affects the selection of probe set for fault localization. We are presenting a novel approach for minimal probe station selection with an objective to reduce probe set size that will be used for fault localization. The probe station selected through this approach results into reduced deployment cost as well as it helps in producing smaller probe set for fault localization reducing overall operation cost. We provide experimental evaluation of the proposed algorithms through simulation results.

Image Captioning Methods and Metrics

Abstract: Image Captioning is one of the emerging topics of research in the field of AI. It uses a combination of Computer Vision (CV) and Natural Language Processing (NLP) to derive features from the image, use this information to identify objects, actions, their relationships, and generate a description for the image. It is most important concept in artificial intelligence applied in the fields like aid to the blind, self-driving cars, and many more. This paper we demonstrates a concise state of art image captioning and its method for caption generation using deep learning concepts. We also determine the approach for image caption generation using Convolutional Neural Network (CNN) and Generative Adversarial Network (GAN) model in deep learning framework. Using this approach system intelligent enough to create sentences for images. It uses the encoder-decoder architecture, where CNN is used for image vector generation and LSTM is used for the generation of a logical sentence using the NLP concepts. Finally, we evaluate the proposed system experimental analysis with numerous existing systems and show the effeteness of system.

Probe station placement algorithm for probe set reduction in network fault detection and localization

Abstract: Fault diagnosis is a central aspect of network fault management. Since faults are unavoidable in communication systems, their quick detection and isolation is essential for the robustness, reliability, and accessibility of a system. Fault localization is done using probing, monitoring, with agents, agent less, through WMI etc. Probing technique for fault localization involves placement of probe stations (Probe stations are specially instrumented nodes from where probes can be sent to monitor the network) which affects the diagnosis capability of the probes sent by the probe stations. Probe station locations affect probing efficiency, monitoring capability, and deployment cost. At the same time, probe station selection affects the selection of probe set for fault localization. We are presenting a novel approach for minimal probe station selection with an objective to reduce probe set size that will be used for fault localization. The probe station selected through this approach results into reduced deployment cost as well as it helps in producing smaller probe set for fault localization reducing overall operation cost. We provide experimental evaluation of the proposed algorithm and provided analysis of same with Shadow Node Reduction (SNR) algorithm through simulation results.

Deep Learning Based Cotton Leaf Disease Detection

Abstract: One of the most widely grown crops in the world, cotton provides a living for many farmers and is essential to the growth of the international economy. However, the actual production of cotton is frequently hampered by a number of problems, with sick cotton leaves ranking highly among them. In order to detect unhealthy cotton leaves, our research uses three distinct deep learning models, including CNN, InceptionV3, and Resnet 152 V2, to categorize cotton leaves or plants as fresh or diseased. The accuracy results for CNN, Inception V3, and Resnet 152V2 are respectively 99.057, 97.170, and 98.113, highlighting the significant contribution these approaches can make to solving this issue.

Software-based fast failure recovery in load balanced SDN-based datacenter networks

Abstract: Load balancing is currently considered as a candidate solution to tackle the emerging problem of increasing bandwidth demand in intra-datacenter networks. Furthermore, because a short disruption of data transfer would corrupt the result of a long procedure of computation, fast failure management mechanisms are considered as integral part of current datacenters. In this paper, we propose a method which uses active probing to detect and manage failures in an OpenFlow based datacenter network exploiting load balancing among equal cost multiple paths. The proposed method is scalable and effective based on the actions it takes without involving the controller in the fast failure recovery procedure.

Design and implementation for automated network troubleshooting using data mining

Abstract: The efficient and effective monitoring of mobile networks is vital given the number of users who rely on such networks and the importance of those networks. The purpose of this paper is to present a monitoring scheme for mobile networks based on the use of rules and decision tree data mining classifiers to upgrade fault detection and handling. The goal is to have optimisation rules that improve anomaly detection. In addition, a monitoring scheme that relies on Bayesian classifiers was also implemented for the purpose of fault isolation and localisation. The data mining techniques described in this paper are intended to allow a system to be trained to actually learn network fault rules. The results of the tests that were conducted allowed for the conclusion that the rules were highly effective to improve network troubleshooting.

Probe station placement algorithm for probe set reduction in network fault localization

Abstract: Fault diagnosis is a central aspect of network fault management. Since faults are unavoidable in communication systems, their quick detection and isolation is essential for the robustness, reliability, and accessibility of a system. Probing technique for fault localization involves placement of probe stations (Probe stations are specially instrumented nodes from where probes can be sent to monitor the network) which affects the diagnosis capability of the probes sent by the probe stations. Probe station locations affect probing efficiency, monitoring capability, and deployment cost. At the same time, probe station selection affects the selection of probe set for fault localization. We are presenting a novel approach for minimal probe station selection with an objective to reduce probe set size that will be used for fault localization. The probe station selected through this approach results into reduced deployment cost as well as it helps in producing smaller probe set for fault localization reducing overall operation cost. We provide experimental evaluation of the proposed algorithms through simulation results.

Network-Wide Initiatives to Control Measurement Mechanisms: A Survey

Abstract: Measurement mechanisms have steadily evolved over the last years and are a key tool for network management. These mechanisms produce results for several network metrics and can be used in different contexts by network administrators. However, the deployment and operation of measurement mechanisms consumes valuable computational and human resources. In this context, approaches to help the administrators to control measurement mechanisms are of paramount importance. In this context, network-wide approaches for such control can provide a larger impact than single device ones. The purpose of this paper is to look at initiatives on the network-wide approaches to control measurement mechanisms in order to provide an integrated perspective. Moreover, we describe criteria that can be used to analyze and compare these initiatives. Furthermore, future trends are discussed in order to predict what the future holds for network-wide control of measurement mechanisms.