Papers by Shyamala Sivakumar
IEEE Transactions on Cybernetics
This paper introduces a discrete-time recurrent neural network architecture using triangular feed... more This paper introduces a discrete-time recurrent neural network architecture using triangular feedback weight matrices that allows a simplified approach to ensuring network and training stability. The triangular structure of the weight matrices is exploited to readily ensure that the eigenvalues of the feedback weight matrix represented by the block diagonal elements lie on the unit circle in the complex z-plane by updating these weights based on the differential of the angular error variable. Such placement of the eigenvalues together with the extended close interaction between state variables facilitated by the nondiagonal triangular elements, enhances the learning ability of the proposed architecture. Simulation results show that the proposed architecture is highly effective in time-series prediction tasks associated with nonlinear and chaotic dynamic systems with underlying oscillatory modes. This modular architecture with dual upper and lower triangular feedback weight matrices mimics fully recurrent network architectures, while maintaining learning stability with a simplified training process. While training, the block-diagonal weights (hence the eigenvalues) of the dual triangular matrices are constrained to the same values during weight updates aimed at minimizing the possibility of overfitting. The dual triangular architecture also exploits the benefit of parsing the input and selectively applying the parsed inputs to the two subnetworks to facilitate enhanced learning performance.
Canadian Journal of Electrical and Computer Engineering
... Shyamala C. Sivakumar is with the Sobey School of Business, Sobey Building, Saint Mary's ... more ... Shyamala C. Sivakumar is with the Sobey School of Business, Sobey Building, Saint Mary's University, 903 Robie Street, Halifax, Nova Scotia B3H ... The amplitude aji of the jet varies slowly with position (i, j) and provides for inherent robustness against translations of the faces ...
Information Fusion, Jul 1, 2011
Clustering techniques have emerged as a popular choice for achieving energy efficiency and scalab... more Clustering techniques have emerged as a popular choice for achieving energy efficiency and scalable performance in large scale sensor networks. Cluster formation is a process whereby sensor nodes decide which cluster head they should associate with among multiple choices. Typically this cluster head selection decision involves a metric based on parameters including residual energy and distance to the cluster head. This decision is a critical embarkation point as a poor choice can lead to increased energy consumption, thus compromising network lifetime. In this paper we present a novel energy efficient cluster formation algorithm based on a multi-criterion optimization technique. Our technique is capable of using multiple individual metrics in the cluster head selection process as input while simultaneously optimizing on the energy efficiency of the individual sensor nodes as well as the overall system. The proposed technique is implemented as a distributed protocol in which each node makes its decision based on local information only. The feasibility of the proposed technique is demonstrated with simulation results. It is shown that the proposed technique outperforms all other well known protocols including LEACH, EECS and HEED resulting in a significant increase in network life.
Proceedings of the 2007 International Conference on Frontiers of High Performance Computing and Networking, 2007
In wireless sensor networks, unbalanced energy dissipation is one of the major causes for hot spo... more In wireless sensor networks, unbalanced energy dissipation is one of the major causes for hot spots. The many-to-one communication pattern used by sensor nodes in most of the data gathering applications leads to such unbalanced energy consumption. Cluster-based protocols attempt to solve this problem by load balancing within the cluster and rotating the job of cluster head every few rounds.
2011 3rd International Congress on Ultra Modern Telecommunications and Control Systems and Workshops, Oct 1, 2011
ABSTRACT The paper presents a dynamic provisioning mechanism through which service providers can ... more ABSTRACT The paper presents a dynamic provisioning mechanism through which service providers can exploit the unused allowable down time of the connections to serve the additional upcoming requests. The algorithm work based on the holding time of the connections and the failure arrival rate over the selected primary or backup paths. The proposed mechanism in this paper routes the requests in a way that any service level agreement (SLA) violation is either avoided or minimized. To achieve this goal, the already established paths are flagged with a newly proposed path metric, path risk factor, to create a memory-aware mechanism of paths' history showing the risk tolerance to SLA violation. This path attribute can be disseminated over the network as a metric of prospective connections. The algorithm takes advantage of the already established connections' history to select the best path regarding the SLA violation with the lowest cost. Simulation results verify that the proposed mechanism has better performance in terms of the blocking rate, the availability satisfaction rate, and the resource utilization than existing algorithms. Performance evaluation is done over two different simulation environments, the network with high link failure arrival rate and the network with normal link failure arrival rate. In addition to the better performance over both network topologies, the algorithm provides more revenue for service providers compared to standard and existing algorithms.
Canadian Conference on Electrical and Computer Engineering 2004 (IEEE Cat. No.04CH37513), 2004
ABSTRACT The core of any quality of service (QoS) scheme is how to monitor and manage traffic flo... more ABSTRACT The core of any quality of service (QoS) scheme is how to monitor and manage traffic flows that have guaranteed QoS. The two QoS approaches used in the literature are, IntServ (integrated service) (White, P.P., 1997) and DiffServ (differentiated service) (Blake, S. et al., 1998). IntServ uses the resource reservation protocol (RSVP) for signaling the path. Decisions on QoS requests are taken by each router independently. DiffServ uses bandwidth brokers as a QoS management model to negotiate requests, communicate with edge nodes and track reservations. We propose an IPv6 QoS management scheme that uses the flow label and traffic class (TC) fields for reserving resources. Edge nodes use these two fields for classification, scheduling and monitoring traffic flows which have requested QoS from the network. Classification is not limited to a number of predefined classes, but on the priority levels (TC field). Processing time is minimized and routing is optimized as routers have to check only the flow identification fields, source IP address and flow label, to direct traffic appropriately. The QoS parameters used are end-to-end delay and packet loss. The simulations are performed on a network simulator (NS-2) (http://www.isi.edu/nsnam/ns/, 2003).
CCECE 2003 - Canadian Conference on Electrical and Computer Engineering. Toward a Caring and Humane Technology (Cat. No.03CH37436), 2003
ABSTRACT With the growth of the Internet and intranets, and the use of multimedia applications, v... more ABSTRACT With the growth of the Internet and intranets, and the use of multimedia applications, video and audio streams, QoS (quality of service) technology has become more relevant and important Yoram Bernet (2000). The next generation Internet protocol (IPv6) provides more features than IPv4 such as more address space and new fields that can be used to enhance and make the usage of IP networks more frequent even with sensitive traffic flows. Current IP networks provide best effort traffic delivery since no QoS features were implemented so other protocols are used to guarantee requests for sensitive traffic flows. IPv6 has implemented two fields that can be used as tools to implement QoS, the two fields are, flow label and traffic class. Flow label is a 20 bit field used by a source to label sequences of packets for which it requests special handling by IPv6 routers. Traffic class is used to indicate the priority level of the traffic flow. IPv6 routers will look at the flow label field plus source address to process the flow packets Silvia Hagen (2002). In this paper, we will investigate IPv6 based end-to-end QoS methodology on quality of service parameters such as end-to-end delay, packet loss and throughput delay using simulations performed with network simulator (NS) Ref.3. The results obtained will be compared with two traditional end-to-end quality of service methods, viz., MPLS (multiprotocol label switching) and RSVP (resource reservation protocol).
Canadian Conference on Electrical and Computer Engineering
The objective of this paper is to recognize speech based on speech prediction techniques using a ... more The objective of this paper is to recognize speech based on speech prediction techniques using a discrete time recurrent neural network (DTRNN) with a block diagonal feedback weight matrix called the block diagonal recurrent neural network (BDRNN). The ability of this network has been investigated for the TIMIT isolated digits spoken by a representative speaker. Simulation results for classifying the utterances show that the size of the BDRNN required is very small compared to multilayer perceptron networks with time delayed feedforward connections
This paper considers the construction method for the block-diagonal recurrent neural network (BDR... more This paper considers the construction method for the block-diagonal recurrent neural network (BDRNN) that is capable of modelling plants with complex eigenvalues. If nonlinear dynamics can be decoupled into a dominant dynamic and several less dominant dynamics, then it is feasible to employ blocks of BDRNNs, one each, to automatically learn each of these dynamics. The advantages of such an approach are that the size of the network is determined automatically and methodically, and faster learning time in comparison with a larger initial network that may be used to learn the overall dynamics. The proposed cascading constructive trajectory learning algorithm constructs a series of BDRNNs whose combined output is required to model the desired dynamic trajectory under consideration. The basic block is directly trained on the desired trajectory being learned, while, each additional cascading block is trained on the residual error between the most recent estimate and the desired trajectory
This paper deals with a discrete time recurrent neural network (DTRNN) with a block-diagonal feed... more This paper deals with a discrete time recurrent neural network (DTRNN) with a block-diagonal feedback weight matrix, called the block-diagonal recurrent neural network (BDRNN), that allows a simplified approach to online trajectory learning. The BDRNN is a sparse but structured architecture in which the feedback connections are restricted to between pairs of state variables. The block-diagonal structure of the BDRNN is exploited to modify the backpropagation-through-time (BPTT) algorithm to reduce the storage requirements while still maintaining exactness and locality of gradient computation. To achieve this, a numerically stable method for recomputing the state variables in the backward pass of the BPTT algorithm is presented
ABSTRACT End-to-end quality of service (QoS) is defined as the capability of a network to provide... more ABSTRACT End-to-end quality of service (QoS) is defined as the capability of a network to provide better service to selected network traffic over various network technologies such as IP, ATM or frame relay networks. Recently users are beginning to combine real-time applications such as voice and video, which have a limited tolerance for network latency and packet loss. To meet these demands, IPv4 was enhanced to handle some quality of service (QoS) aspects such as delay, throughput and packet drop with enhanced methods such as WFQ (weighted fair queuing), RSVP (resource reservation protocol) and MPLS (multiprotocol label switching). This paper uses OPNET (optimum network performance) to simulate a test network with the three QoS methods mentioned and measures parameters such as end-to-end delay, throughput and percentage of traffic lost for four traffic flows with different QoS ratings. This paper investigates the effectiveness of the three QoS methods for traffic flows with varying QoS ratings.
Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering, 2012
The Computer Journal, 2014
Procedia Computer Science, 2014
Software defined networking (SDN) is an emerging approach to handle data forwarding and control s... more Software defined networking (SDN) is an emerging approach to handle data forwarding and control separately. The notion of programmability has central importance in SDN. Two implementation strategies; proprietary and open source, are shaping the trends of the adoptability of SDN by major hardware manufacturers. A group of leading vendors believes that loose coupling between the logical and physical layers of a network hinders the proper provision of physical resources and suggests a proprietary fix to this problem. The other group regards the notion of openness as s key feature of SDN. This paper compares and contrasts these two implementation strategies of SDN by identifying their respective operating principles, features of the product lines, and weakness and strengths.
Electrical and Computer …, 2005
... Shyamala C. Sivakumar is with the Sobey School of Business, Sobey Building, Saint Mary's ... more ... Shyamala C. Sivakumar is with the Sobey School of Business, Sobey Building, Saint Mary's University, 903 Robie Street, Halifax, Nova Scotia B3H ... The amplitude aji of the jet varies slowly with position (i, j) and provides for inherent robustness against translations of the faces ...
IEEE Transactions on Neural Networks, 1999
This paper deals with a discrete-time recurrent neural network (DTRNN) with a block-diagonal feed... more This paper deals with a discrete-time recurrent neural network (DTRNN) with a block-diagonal feedback weight matrix, called the block-diagonal recurrent neural network (BDRNN), that allows a simplified approach to online training and to address network and training stability issues. The structure of the BDRNN is exploited to modify the conventional backpropagation through time (BPTT) algorithm. to reduce its storage requirement by a numerically stable method of recomputing the network state variables. The network and training stability is addressed by exploiting the BDRNN structure to directly monitor and maintain stability during weight updates by developing a functional measure of system stability that augments the cost function being minimized. Simulation results are presented to demonstrate the performance of the BDRNN architecture, its training algorithm, and the stabilization method.
The problem of finding QoS paths involving several combinations of network metrics is NP-complete... more The problem of finding QoS paths involving several combinations of network metrics is NP-complete. This motivates the use of heuristic approaches for finding feasible QoS paths. Many constraint based routing algorithms find QoSpaths by first pruning resources that do ...
Intelligent Quality of …, 2010
Network multimedia applications constitute a large part of Internet traffic and guaranteed delive... more Network multimedia applications constitute a large part of Internet traffic and guaranteed delivery of such traffic is a challenge because of their sensitivity to delay, packet loss and higher bandwidth requirement. The need for guaranteed traffic delivery is exacerbated by the increasing delay experienced by traffic propagating through more than one QoS domain. Hence, there is a need for a flexible and a scalable QoS manager that handles and manages the needs of traffic flows throughout multiple IPv6 domains. The IPv6 QoS manager, presented in this paper, uses a combination of the packets' flow ID and the source address (Domain Global Identifier (DGI)), to process and reserve resources inside an IPv6 domain. To ensure inter-domain QoS management, the QoS domain manager should also communicate with other QoS domains' managers to ensure that traffic flows are guaranteed delivery. In this scheme, the IPv6 QoS manager handles QoS requests by either processing them locally if the intended destination is located locally or forwards the request to the neighboring domain's QoS manager. End-to-end QoS is achieved with an integrated admission and management unit. The feasibility of the proposed QoS management scheme is illustrated for both intra-and inter-domain QoS management. The scalability of the QoS management scheme for inter-domain scenarios is illustrated with simulations for traffic flows propagating through two and three domains. Excellent average end-to-end delay results have been achieved when traffic flow propagates through more than one domain. Simulations show that
Uploads
Papers by Shyamala Sivakumar