Computer Science and Engineering

WiFi mesh networks with outdoor links have become an attractive
option to provide cost-effective broadband connectivity
to rural areas, especially in developing regions. It is
well understood that a TDMA-based approach is necessary
to provide good performance over such networks. While preliminary
prototypes of TDMA-based MAC protocols have
been developed, there is no implementation-based validation/
evaluation in multi-hop settings. In this work, we describe
the elements of a multi-hop MAC implementation
based on the open-source MADWIFI driver. We also present
an evaluation, with a detailed accounting of the various overheads,
on a 4-hop (5-node) path. We show that our implementation
has no system overheads, achieves good throughput,
and low delay/jitter.

Abstract:
Data center interconnectivity is particularly very important and essential for emerging applications such as cloud computing and financial trading. Current data center architectures are built using Ethernet switches or IP routers - both with significant cost and performance deficiencies. We propose for the first time, extending MPLS-TP into the data-center. To this end, a new look-up protocol for MPLS-TP is proposed. Autonomic communication within the data center is possible using our look-up protocol that enables fast creation and deletion of LSPs. The MPLS-TP based data-center that is architected in this paper leads to performance betterments over both IP and Ethernet. To this end, a comprehensive simulations model is also presented. Operations within the data-center using MPLS-TP are also extended to inter-data-center operations using LSP setup across a core network. OAM and performance issues are investigated.

Abstract:
We propose an Impairment Aware RWA algorithm called IA-KS-EDP, which evaluates multiple routing combinations considering the effect of important PLIs using edge-disjoint paths to provision multi-line-rate static traffic demands with minimum wavelengths.

Sensor networks with battery-powered nodes can seldom simultaneously meet the design goals of lifetime, cost, sensing reliability and sensing and transmission coverage. Energy-harvesting, converting ambient energy to electrical energy, has emerged as an alternative to power sensor nodes. By exploiting recharge opportunities and tuning performance parameters based on current and expected energy levels, energy harvesting sensor nodes have the potential to address the conflicting design goals of lifetime and performance. This paper surveys various aspects of energy harvesting sensor systemsarchitecture, energy sources and storage technologies and examples of harvesting-based nodes and applications. The study also discusses the implications of recharge opportunities on sensor node operation and design of sensor network solutions.

The FRACTEL project makes use of long distance wireless links to connect rural and inaccessible areas to a nearby city providing internet, voice and video connectivity. We shall use high gain directional antennas and sectorized antennas for long distance communication over off-the-shelf WiFi hardware. Shorter distance links extend this connectivity to nearby locations. To support long distance links and quality of service requirements, we shall run a TDMA schedule over the FRAC-TEL infrastructure. In this project, we look at the issues of node synchronization which is critical for correct working of the TDMA schedule. Since FRACTEL has a tree topology, we can perform synchronization between two levels at a time and finally get the entire network to know the global clock time. Another aspect of this project is to optimize the power utilization of the nodes. We shall use a proxy based wakeup mechanism to let the main devices sleep while another low power device stays awake and listens for any wakeup calls. When it receives such a signal, it wakes up the main board. In this report, we discuss the steps taken so far in these aspects and the path ahead.

The retail business in India is expanding in leaps and bounds. With increasing competition, each retailer needs to correctly cope up with the impending demand. The retail firms need to take into account various factors including the lead time and the seasonality of goods. In this report, we study the various techniques used by retailers in India to perform demand forecasting and analysis. We present a case study done at Lakewood Malls Pvt. Ltd. where we understand that historical data and judgemental techniques are primarily used for forecasting of demand.

Recent work on bilingual Word Sense Disambiguation (WSD) has shown that a resource deprived language (L 1 ) can benefit from the annotation work done in a resource rich language (L 2 ) via parameter projection. However, this method assumes the presence of sufficient annotated data in one resource rich language which may not always be possible. Instead, we focus on the situation where there are two resource deprived languages, both having a very small amount of seed annotated data and a large amount of untagged data. We then use bilingual bootstrapping, wherein, a model trained using the seed annotated data of L 1 is used to annotate the untagged data of L 2 and vice versa using parameter projection. The untagged instances of L 1 and L 2 which get annotated with high confidence are then added to the seed data of the respective languages and the above process is repeated. Our experiments show that such a bilingual bootstrapping algorithm when evaluated on two different domains with small seed sizes using Hindi (L 1 ) and Marathi (L 2 ) as the language pair performs better than monolingual bootstrapping and significantly reduces annotation cost.

Several bilingual WSD algorithms which exploit translation correspondences between parallel corpora have been proposed. However, the availability of such parallel corpora itself is a tall task for some of the resource constrained languages of the world. We propose an unsupervised bilingual EM based algorithm which relies on the counts of translations to estimate sense distributions. No parallel or sense annotated corpora are needed. The algorithm relies on a synset-aligned bilingual dictionary and in-domain corpora from the two languages. A symmetric generalized Expectation Maximization formulation is used wherein the sense distributions of words in one language are estimated based on the raw counts of the words in the aligned synset in the target language. The overall performance of our algorithm when tested on 4 language-domain pairs is better than current state-of-the-art knowledge based and bilingual unsupervised approaches.

Wireless sensor networks (WSNs) consist of a large number of small, battery-powered wireless sensor nodes and thereby constrained with energy resource. The network as a whole must minimize the energy usage in order to enable untethered and unattended operation for an extended period of time. One fundamental way of reducing such energy usage and enhancing lifetime is judicial placement of sensor nodes within the network area. The present work proposes a lifetime-enhancing node-deployment strategy. Primarily more nodes are deployed towards the sink exploiting the fact that more energy is drained out from the nodes close to the sink. Further, locations which are approximately equi-distant from the sink are assigned different classes of priorities according to their responsibility of forwarding data of nodes located in the neighbouring cells. The principle of this strategy is justified by Lemma and corresponding proof. Exhaustive simulation is conducted to observe the impact of node distribution on network lifetime. The result is compared with one of the existing deployment strategies, which shows our scheme outperforms the existing one in terms of network lifetime.

A robust and efficient algorithm to recognize handwritten Bangla (Bengali) characters in machine-printed forms is proposed. It is based on the combination of gradient features and Haar wavelet coefficients. The gradient feature is used to capture local characteristics, and for its sensitivity to the usual deformation and idiosyncrasy of handwritten characters, wavelet transform is used for multi-resolution analysis of character images. Such a strategy with combined features captures adequate global characteristics in different scales. Two feature-combination schemes are devised and tested on test images of 4372 instances of 49 characters and 10 numerals, after being trained by a set of 59×25 = 1475 images. Finally, a k-NN classifier is used for the character recognition, which shows 87.65% and 88.95% recognition accuracies for the two schemes.

Recent advances in wireless communication, low power sensors and microcontrollers enable the deployment of large-scale wireless sensor networks. A fundamental problem in wireless sensor networks is localization i.e. determination of geographical locations of sensor nodes. This paper addresses the problem of location discovery of the nodes in wireless sensor network. We propose a localization technique by means of which a sensor node can determine its location with enhanced accuracy by listening wireless transmissions from three or more beacons. The proposed method is based on RSSI technique that does not require any additional complexity in construction of the sensor nodes. Performance of the proposed localization scheme is compared with one of the existing schemes. Qualitative analysis is done considering communication, storage and computational overhead as performance metrics, whereas quantitative analysis is done considering localization error as a performance metric. The results confirm our scheme's supremacy over the competing scheme.

Despite the apparent randomness of the Internet, we discover some surprisingly simple power-laws of the Internet topology. These power-laws hold for three snapshots of the Internet, between November 1997 and December 1998, despite a 45% growth of its size during that period. We show that our power-laws t the real data very well resulting in correlation coe cients of 96% or higher.

Recent work has shown that the node degree in the WWW induced graph and the AS-level Internet topology exhibit power laws. Since then several algorithms have been proposed to generate such power law graphs. In this paper we evaluate the effectiveness of these generators to generate representative AS-level topologies. Our conclusions are mixed. Although they (mostly) do a reasonable job at capturing the power law exponent, they do less well in capturing the clustering phenomena exhibited by the Internet topology. Based on these results we propose a variation of the recent incremental topology generator of [6] that is more successful at matching the power law exponent and the clustering behavior of the Internet. Last, we comment on the small world behavior of the Internet topology.

In large and constantly evolving networks, it is difficult to determine how the network is actually laid out. Yet this information is invaluable for network management, simulation, and server siting. Traditional topology discovery algorithms are based on SNMP, which is not universally deployed. We describe several heuristics and algorithms to discover both intra-domain and Internet backbone topology while making as few assumptions about the network as possible. We quantitatively evaluate their performance and also present a new technique for visualizing Internet backbone topology.

One central issue in the formal design and analysis of reactive systems is the notion of refinement that asks whether all behaviors of the implementation is allowed by the specification. The local interpretation of behavior leads to the notion of simulation. Alternating transition systems (ATSs) provide a general model for composite reactive systems, and the simulation relation for ATSs is known as alternating simulation. The simulation relation for fair transition systems is called fair simulation. In this work our main contributions are as follows: (1) We present an improved algorithm for fair simulation with Büchi fairness constraints; our algorithm requires O(n 3 · m) time as compared to the previous known O(n 6 )-time algorithm, where n is the number of states and m is the number of transitions.

We investigate a model-theoretic property that generalizes the classical notion of "preservation under substructures". We call this property preservation under substructures modulo bounded cores, and present a syntactic characterization via Σ 0 2 sentences for properties of arbitrary structures definable by FO sentences. As a sharper characterization, we further show that the count of existential quantifiers in the Σ 0 2 sentence equals the size of the smallest bounded core. We also present our results on the sharper characterization for special fragments of FO and also over special classes of structures. We present a (not FO-definable) class of finite structures for which the sharper characterization fails, but for which the classical Łoś-Tarski preservation theorem holds. As a fallout of our studies, we obtain combinatorial proofs of the Łoś-Tarski theorem for some of the aforementioned cases.

Identification of protein families is a computational biology challenge that needs efficient and reliable methods. Here we introduce the concept of dominance and propose a novel combined approach based on Distance Alignment Search Tool (DAST), which contains an exact algorithm with bounds. Our experiments show that this method successfully finds the most similar proteins in a set without solving all instances.