arXiv subjects

Akshay Jain

Publications and source records attributed to Akshay Jain.

27 records · Page 2Linked to original sources

arXiv2021

Computing Graph Neural Networks: A Survey from Algorithms to Accelerators

Graph Neural Networks (GNNs) have exploded onto the machine learning scene in recent years owing to their capability to model and learn from graph-structured data. Such an ability has strong implications in a wide variety of fields whose data is inherently relational, for which conventional neural networks do not perform well. Indeed, as recent reviews can attest, research in the area of GNNs has grown rapidly and has lead to the development of a variety of GNN algorithm variants as well as to the exploration of groundbreaking applications in chemistry, neurology, electronics, or communication networks, among others. At the current stage of research, however, the efficient processing of GNNs is still an open challenge for several reasons. Besides of their novelty, GNNs are hard to compute due to their dependence on the input graph, their combination of dense and very sparse operations, or the need to scale to huge graphs in some applications. In this context, this paper aims to make two main contributions. On the one hand, a review of the field of GNNs is presented from the perspective of computing. This includes a brief tutorial on the GNN fundamentals, an overview of the evolution of the field in the last decade, and a summary of operations carried out in the multiple phases of different GNN algorithm variants. On the other hand, an in-depth analysis of current software and hardware acceleration schemes is provided, from which a hardware-software, graph-aware, and communication-centric vision for GNN accelerators is distilled.

cs.LG↗

arXiv2021

Characterizing the Communication Requirements of GNN Accelerators: A Model-Based Approach

Relational data present in real world graph representations demands for tools capable to study it accurately. In this regard Graph Neural Network (GNN) is a powerful tool, wherein various models for it have also been developed over the past decade. Recently, there has been a significant push towards creating accelerators that speed up the inference and training process of GNNs. These accelerators, however, do not delve into the impact of their dataflows on the overall data movement and, hence, on the communication requirements. In this paper, we formulate analytical models that capture the amount of data movement in the most recent GNN accelerator frameworks. Specifically, the proposed models capture the dataflows and hardware setup of these accelerator designs and expose their scalability characteristics for a set of hardware, GNN model and input graph parameters. Additionally, the proposed approach provides means for the comparative analysis of the vastly different GNN accelerators.

cs.AR↗

arXiv2020

Radiation pattern prediction for Metasurfaces: A Neural Network based approach

As the current standardization for the 5G networks nears completion, work towards understanding the potential technologies for the 6G wireless networks is already underway. One of these potential technologies for the 6G networks are Reconfigurable Intelligent Surfaces (RISs). They offer unprecedented degrees of freedom towards engineering the wireless channel, i.e., the ability to modify the characteristics of the channel whenever and however required. Nevertheless, such properties demand that the response of the associated metasurface (MSF) is well understood under all possible operational conditions. While an understanding of the radiation pattern characteristics can be obtained through either analytical models or full wave simulations, they suffer from inaccuracy under certain conditions and extremely high computational complexity, respectively. Hence, in this paper we propose a novel neural networks based approach that enables a fast and accurate characterization of the MSF response. We analyze multiple scenarios and demonstrate the capabilities and utility of the proposed methodology. Concretely, we show that this method is able to learn and predict the parameters governing the reflected wave radiation pattern with an accuracy of a full wave simulation (98.8%-99.8%) and the time and computational complexity of an analytical model. The aforementioned result and methodology will be of specific importance for the design, fault tolerance and maintenance of the thousands of RISs that will be deployed in the 6G network environment.

eess.SP↗

arXiv2020

Are Mobility Management Solutions Ready for 5G and Beyond?

Enabling users to move to different geographical locations within a network and still be able to maintain their connectivity and most essentially, continuity of service, is what makes any wireless network ubiquitous. Whilst challenging, modern day wireless networks, such as 3GPP-LTE, provision satisfactory mobility management (MM) performance. However, it is estimated that the number of mobile subscriptions will approximately touch 9 billion and the amount of data traffic will expand by 5 times in 2024 as compared to 2018. Further, it is expected that this trend of exponential growth will be maintained well into the future. To cope with such an exponential increase in cellular traffic and users alongside a burgeoning demand for higher Quality of Service (QoS), the future networks are expected to be highly dense and heterogeneous. This will severely challenge the existing MM solutions and ultimately render them ineffective as they will not be able to provide the required reliability, flexibility, and scalability. Consequently, to serve the 5G and beyond 5G networks, a new perspective to MM is required. Hence, in this article we present a novel discussion of the functional requirements from MM strategies for these networks. We present a novel qualitative assessment of the state-of-the-art mechanisms based on their ability to satisfy the reliability, flexibility and scalability criteria for future MM strategies. We then present a study detailing the research challenges that exist in the design and implementation of MM strategies for 5G and beyond networks. Further, we chart out the potential MM solutions and the associated capabilities they offer to tackle the persistent challenges. We conclude this paper with a vision for the 5G and beyond MM mechanisms.

cs.NI↗

arXiv2020

User Association and Resource Allocation in 5G (AURA-5G): A Joint Optimization Framework

In this paper, we provide a novel application aware user association and resource allocation framework, i.e., AURA-5G, which utilizes a joint optimization strategy to accomplish the same. Concretely, our methodology considers all the real network constraints that will be prevalent in the 5G networks as well as practical deployment scenarios. Furthermore, AURA-5G, being an application aware framework, considers the resource requirements of both eMBB and mMTC services whilst performing the optimization task. We have demonstrated that our strategy performs significantly better than the baseline algorithm, given any of the multiple combinations of network constraints explored in this paper. In addition, we have also presented a novel computational complexity analysis for the AURA-5G framework as well as a solvability and convergence time analysis. Such an analysis will be beneficial for both industry and academia in determining the applicability and performance of the AURA-5G framework, given the scenario and constraints. Lastly, we have also provisioned a short study on the aspect of network re-dimensioning, wherein we demonstrate the efficacy of the AURA-5G framework in providing insights to the operators with regards to their deployment and how they can utilize it to optimize the performance of their networks.

cs.NI↗

arXiv2020

ManyModalQA: Modality Disambiguation and QA over Diverse Inputs

We present a new multimodal question answering challenge, ManyModalQA, in which an agent must answer a question by considering three distinct modalities: text, images, and tables. We collect our data by scraping Wikipedia and then utilize crowdsourcing to collect question-answer pairs. Our questions are ambiguous, in that the modality that contains the answer is not easily determined based solely upon the question. To demonstrate this ambiguity, we construct a modality selector (or disambiguator) network, and this model gets substantially lower accuracy on our challenge set, compared to existing datasets, indicating that our questions are more ambiguous. By analyzing this model, we investigate which words in the question are indicative of the modality. Next, we construct a simple baseline ManyModalQA model, which, based on the prediction from the modality selector, fires a corresponding pre-trained state-of-the-art unimodal QA model. We focus on providing the community with a new manymodal evaluation set and only provide a fine-tuning set, with the expectation that existing datasets and approaches will be transferred for most of the training, to encourage low-resource generalization without large, monolithic training sets for each new task. There is a significant gap between our baseline models and human performance; therefore, we hope that this challenge encourages research in end-to-end modality disambiguation and multimodal QA models, as well as transfer learning. Code and data available at: https://github.com/hannandarryl/ManyModalQA

cs.CL↗

arXiv2017

Joint Access-Backhaul Perspective on Mobility Management in 5G Networks

The ongoing efforts in the research development and standardization of 5G, by both industry and academia, have resulted in the identification of enablers (Software Defined Networks, Network Function Virtualization, Distributed Mobility Management, etc.) and critical areas (Mobility management, Interference management, Joint access-backhaul mechanisms, etc.) that will help achieve the 5G objectives. During these efforts, it has also been identified that the 5G networks due to their high degree of heterogeneity, high QoS demand and the inevitable density (both in terms of access points and users), will need to have efficient joint backhaul and access mechanisms as well as enhanced mobility management mechanisms in order to be effective, efficient and ubiquitous. Therefore, in this paper we first provide a discussion on the evolution of the backhaul scenario, and the necessity for joint access and backhaul optimization. Subsequently, and since mobility management mechanisms can entail the availability, reliability and heterogeneity of the future backhaul/fronthaul networks as parameters in determining the most optimal solution for a given context, a study with regards to the effect of future backhaul/fronthaul scenarios on the design and implementation of mobility management solutions in 5G networks has been performed.

cs.NI↗

arXiv2017

Mobility Management as a Service for 5G Networks

Mobility Management (MM) techniques have conventionally been centralized in nature, wherein a single network entity has been responsible for handling the mobility related tasks of the mobile nodes attached to the network. However, an exponential growth in network traffic and the number of users has ushered in the concept of providing Mobility Management as a Service (MMaaS) to the wireless nodes attached to the 5G networks. Allowing for on-demand mobility management solutions will not only provide the network with the flexibility that it needs to accommodate the many different use cases that are to be served by future networks, but it will also provide the network with the scalability that is needed alongside the flexibility to serve future networks. And hence, in this paper, a detailed study of MMaaS has been provided, highlighting its benefits and challenges for 5G networks. Additionally, the very important property of granularity of service which is deeply intertwined with the scalability and flexibility requirements of the future wireless networks, and a consequence of MMaaS, has also been discussed in detail.

cs.NI↗

arXiv2013

Dynamic Clustering Protocol for Data Forwarding in Wireless Sensor Networks

Energy being the very key concern area with sensor networks, so the main focus lies in developing a mechanism to increase the lifetime of a sensor network by energy balancing. To achieve energy balancing and maximizing network lifetime we use an idea of clustering and dividing the whole network into different clusters. In this paper we propose a dynamic cluster formation method where clusters are refreshed periodically based on residual energy, distance and cost. Refreshing clustering minimizes workload of any single node and in turn enhances the energy conservation. Sleep and wait methodology is applied to the proposed protocol to enhance the network lifetime by turning the nodes on and off according to their duties. The node that has some data to be transmitted is in on state and after forwarding its data to the cluster head it changes its state to off which saves the energy of entire network. Simulations have been done using MAT lab. Simulation results prove the betterment of our proposed method over the existing Leach protocol.

cs.NI↗