-
Memory Efficient and Staleness Free Pipeline Parallel DNN Training Framework with Improved Convergence Speed
Authors:
Ankita Dutta,
Nabendu Chaki,
Rajat K. De
Abstract:
High resource requirement for Deep Neural Network (DNN) training across multiple GPUs necessitates development of various parallelism techniques. In this paper, we introduce two interconnected DNN training frameworks, namely, V-TiMePReSt and I-TiMePReSt, based on pipeline parallelism, a variant of model parallelism. V-TiMePReSt is a completely staleness-free system which enables the DNNs to be tra…
▽ More
High resource requirement for Deep Neural Network (DNN) training across multiple GPUs necessitates development of various parallelism techniques. In this paper, we introduce two interconnected DNN training frameworks, namely, V-TiMePReSt and I-TiMePReSt, based on pipeline parallelism, a variant of model parallelism. V-TiMePReSt is a completely staleness-free system which enables the DNNs to be trained on the latest updated weights in each stage of all forward and backward passes. Developing staleness-aware systems at the expense of weight stashing reduces GPU-memory consumption, however, increases the number of epochs to converge. Thus, we introduce I-TiMePReSt, which is also a staleness-aware system, but not at the expense of weight stashing. It does not rely solely on the stale weights or the latest updated weights. I-TiMePReSt computes an intermediate weight towards the latter and performs backward pass on it. Additionally, we formulate the significance of the stale weights mathematically depending on the degree of staleness. In contrast to V-TiMePReSt, I-TiMePReSt works based on the assumption that stale weights have a significant contribution in training, which can be quantified mathematically based on the degree of staleness, although there are other contributory factors which should not be ignored. Experimental results show that V-TiMePReSt is advantageous over existing models in terms of $1)$ the extent of staleness of the weight parameter values and $2)$ GPU memory efficiency, while I-TiMePReSt is superior in terms of $1)$ removing staleness of the weight parameters without removing weight stashing and $2)$ maintaining the trade-off between GPU memory consumption and convergence speed (number of epochs).
△ Less
Submitted 27 September, 2025;
originally announced September 2025.
-
TiMePReSt: Time and Memory Efficient Pipeline Parallel DNN Training with Removed Staleness
Authors:
Ankita Dutta,
Nabendu Chaki,
Rajat K. De
Abstract:
DNN training is time-consuming and requires efficient multi-accelerator parallelization, where a single training iteration is split over available accelerators. Current approaches often parallelize training using intra-batch parallelization. Combining inter-batch and intra-batch pipeline parallelism is common to further improve training throughput. In this article, we develop a system, called TiMe…
▽ More
DNN training is time-consuming and requires efficient multi-accelerator parallelization, where a single training iteration is split over available accelerators. Current approaches often parallelize training using intra-batch parallelization. Combining inter-batch and intra-batch pipeline parallelism is common to further improve training throughput. In this article, we develop a system, called TiMePReSt, that combines them in a novel way which helps to better overlap computation and communication, and limits the amount of communication. The traditional pipeline-parallel training of DNNs maintains similar working principle as sequential or conventional training of DNNs by maintaining consistent weight versions in forward and backward passes of a mini-batch. Thus, it suffers from high GPU memory footprint during training. In this paper, experimental study demonstrates that compromising weight consistency doesn't decrease prediction capability of a parallelly trained DNN. Moreover, TiMePReSt overcomes GPU memory overhead and achieves zero weight staleness. State-of-the-art techniques often become costly in terms of training time. In order to address this issue, TiMePReSt introduces a variant of intra-batch parallelism that parallelizes the forward pass of each mini-batch by decomposing it into smaller micro-batches. A novel synchronization method between forward and backward passes reduces training time in TiMePReSt. The occurrence of multiple sequence problem and its relation with version difference have been observed in TiMePReSt. This paper presents a mathematical relationship between the number of micro-batches and worker machines, highlighting the variation in version difference. A mathematical expression has been developed to calculate version differences for various combinations of these two without creating diagrams for all combinations.
△ Less
Submitted 23 October, 2024; v1 submitted 18 October, 2024;
originally announced October 2024.
-
An Annexure to the Paper "Driving the Technology Value Stream by Analyzing App Reviews"
Authors:
Souvick Das,
Novarun Deb,
Agostino Cortesi,
Nabendu Chaki
Abstract:
This paper presents a novel framework that utilizes Natural Language Processing (NLP) techniques to understand user feedback on mobile applications. The framework allows software companies to drive their technology value stream based on user reviews, which can highlight areas for improvement. The framework is analyzed in depth, and its modules are evaluated for their effectiveness. The proposed ap…
▽ More
This paper presents a novel framework that utilizes Natural Language Processing (NLP) techniques to understand user feedback on mobile applications. The framework allows software companies to drive their technology value stream based on user reviews, which can highlight areas for improvement. The framework is analyzed in depth, and its modules are evaluated for their effectiveness. The proposed approach is demonstrated to be effective through an analysis of reviews for sixteen popular Android Play Store applications over a long period of time.
△ Less
Submitted 8 March, 2023;
originally announced March 2023.
-
AFSCR: Annotation of Functional Satisfaction Conditions and their Reconciliation within i* models
Authors:
Novarun Deb,
Nabendu Chaki
Abstract:
Context: Researchers, both in industry and academia, are facing the challenge of leveraging the benefits of goal oriented requirements engineering (GORE) techniques to business compliance management. This requires analyzing goal models along with their semantics. However, most prominent goal modeling frameworks have no means of capturing the semantics of goals (except what is trivially conveyed by…
▽ More
Context: Researchers, both in industry and academia, are facing the challenge of leveraging the benefits of goal oriented requirements engineering (GORE) techniques to business compliance management. This requires analyzing goal models along with their semantics. However, most prominent goal modeling frameworks have no means of capturing the semantics of goals (except what is trivially conveyed by their nomenclature).
Objective: In this paper, we propose the Annotation of Functional Satisfaction Conditions and their Reconciliation (AFSCR) framework for doing the same. The entire framework is presented with respect to i* modeling constructs.
Method: This is a semi-automated framework that requires analysts to annotate individual goals with their immediate goal satisfaction conditions. The AFSCR framework can then reconcile these satisfaction conditions for every goal and verify whether the derived set of cumulative satisfaction conditions is in harmony with the intended set of goal satisfaction conditions.
Result: If the derived and intended sets of satisfaction conditions are in conflict, the framework raises entailment and/or consistency flags. Whenever a conflict is flagged, the framework also provides alternate solutions and possible workaround strategies to the analysts by refactoring the given i* model.
Conclusion: In this paper we present a new framework that uses satisfaction conditions for going beyond the nomenclature and capturing the functional semantics of the goals within i* models. The analysis performed during the reconciliation process is generic enough and can be adapted to any goal modeling framework if required.
△ Less
Submitted 12 May, 2019;
originally announced May 2019.
-
Extracting State Transition Models from i* Models
Authors:
Novarun Deb,
Nabendu Chaki,
Aditya Ghose
Abstract:
i* models are inherently sequence agnostic. There is an immediate need to bridge the gap between such a sequence agnostic model and an industry implemented process modelling standard like Business Process Modelling Notation (BPMN). This work is an attempt to build State Transition Models from i* models. In this paper, we first spell out the Naive Algorithm formally, which is on the lines of Formal…
▽ More
i* models are inherently sequence agnostic. There is an immediate need to bridge the gap between such a sequence agnostic model and an industry implemented process modelling standard like Business Process Modelling Notation (BPMN). This work is an attempt to build State Transition Models from i* models. In this paper, we first spell out the Naive Algorithm formally, which is on the lines of Formal Tropos. We demonstrate how the growth of the State Transition Model Space can be mapped to the problem of finding the number of possible paths between the Least Upper Bound (LUB) and the Greatest Lower Bound (GLB) of a k-dimensional hypercube Lattice structure. We formally present the mathematics for doing a quantitative analysis of the space growth. The Naive Algorithm has its main drawback in the hyperexponential explosion caused in the State Transition Model space. This is identified and the Semantic Implosion Algorithm is proposed which exploits the temporal information embedded within the i* model of an enterprise to reduce the rate of growth of the State Transition Model space. A comparative quantitative analysis between the two approaches concludes the superiority of the Semantic Implosion Algorithm.
△ Less
Submitted 24 July, 2015;
originally announced July 2015.
-
Study of Security Issues in Pervasive Environment of Next Generation Internet of Things
Authors:
Tapalina Bhattasali,
Rituparna Chaki,
Nabendu Chaki
Abstract:
Internet of Things is a novel concept that semantically implies a world-wide network of uniquely addressable interconnected smart objects. It is aimed at establishing any paradigm in computing. This environment is one where the boundary between virtual and physical world is eliminated. As the network gets loaded with hitherto unknown applications, security threats also become rampant. Current secu…
▽ More
Internet of Things is a novel concept that semantically implies a world-wide network of uniquely addressable interconnected smart objects. It is aimed at establishing any paradigm in computing. This environment is one where the boundary between virtual and physical world is eliminated. As the network gets loaded with hitherto unknown applications, security threats also become rampant. Current security solutions fail as new threats appear to de-struct the reliability of information. The network has to be transformed to IPv6 enabled network to address huge number of smart objects. Thus new addressing schemes come up with new attacks. Real time analysis of information from the heterogeneous smart objects needs use of cloud services. This can fall prey to cloud specific security threats. Therefore need arises for a review of security threats for a new area having huge demand. Here a study of security issues in this domain is briefly presented.
△ Less
Submitted 19 June, 2014;
originally announced June 2014.
-
Ant Colony based Routing for Mobile Ad-Hoc Networks towards Improved Quality of Services
Authors:
Bibhas Roy,
Suman Banik,
Parthi Dey,
Sugata Sanyal,
Nabendu Chaki
Abstract:
Mobile Ad Hoc Network (MANET) is a dynamic multihop wireless network which is established by a set of mobile nodes on a shared wireless channel. One of the major issues in MANET is routing due to the mobility of the nodes. Routing means the act of moving information across an internet work from a source to a destination. When it comes to MANET, the complexity increases due to various characteristi…
▽ More
Mobile Ad Hoc Network (MANET) is a dynamic multihop wireless network which is established by a set of mobile nodes on a shared wireless channel. One of the major issues in MANET is routing due to the mobility of the nodes. Routing means the act of moving information across an internet work from a source to a destination. When it comes to MANET, the complexity increases due to various characteristics like dynamic topology, time varying QoS requirements, limited resources and energy etc. QoS routing plays an important role for providing QoS in wireless ad hoc networks. The most complex issue in this kind of networks is to find a path between the communication end points satisfying QoS requirement for the user. Nature-inspired algorithms (swarm intelligence) such as ant colony optimization ACO)algorithms have shown to be a good technique for developing routing algorithms for MANETs.
In this paper, a new QoS algorithm for mobile ad hoc network has been proposed. The proposed algorithm combines the idea of Ant Colony Optimization (ACO) with Optimized Link State Routing (OLSR) protocol to identify multiple stable paths between source and destination nodes.
△ Less
Submitted 10 December, 2013; v1 submitted 7 December, 2013;
originally announced December 2013.
-
A New Scale for Attribute Dependency in Large Database Systems
Authors:
Soumya Sen,
Anjan Dutta,
Agostino Cortesi,
Nabendu Chaki
Abstract:
Large, data centric applications are characterized by its different attributes. In modern day, a huge majority of the large data centric applications are based on relational model. The databases are collection of tables and every table consists of numbers of attributes. The data is accessed typically through SQL queries. The queries that are being executed could be analyzed for different types of…
▽ More
Large, data centric applications are characterized by its different attributes. In modern day, a huge majority of the large data centric applications are based on relational model. The databases are collection of tables and every table consists of numbers of attributes. The data is accessed typically through SQL queries. The queries that are being executed could be analyzed for different types of optimizations. Analysis based on different attributes used in a set of query would guide the database administrators to enhance the speed of query execution. A better model in this context would help in predicting the nature of upcoming query set. An effective prediction model would guide in different applications of database, data warehouse, data mining etc. In this paper, a numeric scale has been proposed to enumerate the strength of associations between independent data attributes. The proposed scale is built based on some probabilistic analysis of the usage of the attributes in different queries. Thus this methodology aims to predict future usage of attributes based on the current usage.
△ Less
Submitted 27 June, 2012;
originally announced June 2012.
-
Learners' Quanta based Design of a Learning Management System
Authors:
Souvik Sengupta,
Nabendu Chaki,
Ranjan Dasgupta
Abstract:
In this paper IEEE Learning Technology System Architecture (LTSA) for LMS software has been analyzed. It has been observed that LTSA is too abstract to be adapted in a uniform way by LMS developers. A Learners' Quanta based high level design that satisfies the IEEE LTSA standard has been proposed for future development of efficient LMS software. A hybrid model of learning fitting into LTSA model h…
▽ More
In this paper IEEE Learning Technology System Architecture (LTSA) for LMS software has been analyzed. It has been observed that LTSA is too abstract to be adapted in a uniform way by LMS developers. A Learners' Quanta based high level design that satisfies the IEEE LTSA standard has been proposed for future development of efficient LMS software. A hybrid model of learning fitting into LTSA model has also been proposed while designing.
△ Less
Submitted 19 January, 2012;
originally announced January 2012.
-
A Stochastic Net Model for Controlling Bullwhip Effect in Virtual Multi-Tier Retail Network
Authors:
Bidyut Biman Sarkar,
Sugata Sanyal,
Nabendu Chaki
Abstract:
Supply Chain operation is an integrated business process starting from primary supplier to end user and the process produce products, services and information. A successful chain will explore technology, lean operations, and quality management by adding value for customers and stakeholders. It is a strategic alliance among the partnering enterprises without geographical boundary. Every chain has i…
▽ More
Supply Chain operation is an integrated business process starting from primary supplier to end user and the process produce products, services and information. A successful chain will explore technology, lean operations, and quality management by adding value for customers and stakeholders. It is a strategic alliance among the partnering enterprises without geographical boundary. Every chain has its own unique set of market demands and operating challenges. Retailing is one such service domain of Supply Chain vulnerable to bullwhip effects. Demand uncertainty is one of the root causes of Bullwhip effects. This paper calls for modeling of a demand driven multi-tier stochastic Retail Chain to work against the Bullwhip effect. The proposed model of the operational chain will ensure significant return of share to the retailer through the sophisticated transaction management, real-time inventory management and the ability to track all inventory movements.
△ Less
Submitted 2 November, 2011; v1 submitted 2 November, 2011;
originally announced November 2011.
-
A Survey on Security Issues in Cloud Computing
Authors:
Rohit Bhadauria,
Rituparna Chaki,
Nabendu Chaki,
Sugata Sanyal
Abstract:
Cloud Computing holds the potential to eliminate the requirements for setting up of high-cost computing infrastructure for the IT-based solutions and services that the industry uses. It promises to provide a flexible IT architecture, accessible through internet for lightweight portable devices. This would allow many-fold increase in the capacity or capabilities of the existing and new software. In…
▽ More
Cloud Computing holds the potential to eliminate the requirements for setting up of high-cost computing infrastructure for the IT-based solutions and services that the industry uses. It promises to provide a flexible IT architecture, accessible through internet for lightweight portable devices. This would allow many-fold increase in the capacity or capabilities of the existing and new software. In a cloud computing environment, the entire data reside over a set of networked resources, enabling the data to be accessed through virtual machines. Since these data centers may lie in any corner of the world beyond the reach and control of users, there are multifarious security and privacy challenges that need to be understood and taken care of. Also, one can never deny the possibility of a server breakdown that has been witnessed, rather quite often in the recent times. There are various issues that need to be dealt with respect to security and privacy in a cloud computing scenario. This extensive survey paper aims to elaborate and analyze the numerous unresolved issues threatening the Cloud computing adoption and diffusion affecting the various stake-holders linked to it.
△ Less
Submitted 21 May, 2013; v1 submitted 25 September, 2011;
originally announced September 2011.
-
A State-of-the-art Survey on IDS for Mobile Ad-Hoc Networks and Wireless Mesh Networks
Authors:
Novarun Deb,
Manali Chakraborty,
Nabendu Chaki
Abstract:
An Intrusion Detection System (IDS) detects malicious and selfish nodes in a network. Ad hoc networks are often secured by using either intrusion detection or by secure routing. Designing efficient IDS for wireless ad-hoc networks that would not affect the performance of the network significantly is indeed a challenging task. Arguably, the most common thing in a review paper in the domain of wirel…
▽ More
An Intrusion Detection System (IDS) detects malicious and selfish nodes in a network. Ad hoc networks are often secured by using either intrusion detection or by secure routing. Designing efficient IDS for wireless ad-hoc networks that would not affect the performance of the network significantly is indeed a challenging task. Arguably, the most common thing in a review paper in the domain of wireless networks is to compare the performances of different solutions using simulation results. However, variance in multiple configuration aspects including that due to different underlying routing protocols, makes the task of simulation based comparative evaluation of IDS solutions somewhat unrealistic. In stead, the authors have followed an analytic approach to identify the gaps in the existing IDS solutions for MANETs and wireless mesh networks. The paper aims to ease the job of a new researcher by exposing him to the state of the art research issues on IDS. Nearly 80% of the works cited in this paper are published with in last 3 to 4 years.
△ Less
Submitted 30 August, 2011;
originally announced August 2011.
-
Quality Evaluation of Conceptual Level Object Multidimensional Data Model
Authors:
Supriya Chakrabarty,
Nabendu Chaki
Abstract:
The advancement of technology facilitates explosive growth of mobile usage in the last decade. Numerous applications have been developed to support its usage. However, gap in technology exists in obtaining correct and trusted values for evaluation indexes of the precise amount of usage. The claims of loss in revenue by the service providers could be more due to unexpected behaviour of the hardware…
▽ More
The advancement of technology facilitates explosive growth of mobile usage in the last decade. Numerous applications have been developed to support its usage. However, gap in technology exists in obtaining correct and trusted values for evaluation indexes of the precise amount of usage. The claims of loss in revenue by the service providers could be more due to unexpected behaviour of the hardware. A similar mistrust is often observed in the users of the services. A trustworthy subscription scheme is in demand for consumers whereas revenue needs to be assured of the service providers. Multiple Authorizations by Multiple Owners (MAMO) has already been introduced as a technology to build trust in the third party billing system. In this paper, MAMO is extended to ensure trustworthiness of the parameters for subscription. Along with call transaction data are reconciled to assure the proper revenue generation.
△ Less
Submitted 27 August, 2011;
originally announced August 2011.
-
Modeling Smart Grid using Generalized Stochastic Petri Net
Authors:
Amrita Dey,
Nabendu Chaki,
Sugata Sanyal
Abstract:
Building smart grid for power system is a major challenge for safe, automated and energy efficient usage of electricity. The full implementation of the smart grid will evolve over time. However, before a new set of infrastructures are invested to build the smart grid, proper modeling and analysis is needed to avoid wastage of resources. Modeling also helps to identify and prioritize appropriate sy…
▽ More
Building smart grid for power system is a major challenge for safe, automated and energy efficient usage of electricity. The full implementation of the smart grid will evolve over time. However, before a new set of infrastructures are invested to build the smart grid, proper modeling and analysis is needed to avoid wastage of resources. Modeling also helps to identify and prioritize appropriate systems parameters. In this paper, an all comprehensive model of smart grid have been proposed using Generalized Stochastic Petri Nets (GSPN). The model is used to analyze the constraints and deliverables of the smart power grid of future.
△ Less
Submitted 20 August, 2011;
originally announced August 2011.
-
QoS Routing using OLSR with Optimization for Flooding
Authors:
Suman Banik,
Bibhash Roy,
Parthi Dey,
Nabendu Chaki,
Sugata Sanyal
Abstract:
Mobile Ad-hoc Network (MANET) is the self organizing collection of mobile nodes. The communication in MANET is done via a wireless media. Ad hoc wireless networks have massive commercial and military potential because of their mobility support. Due to demanding real time multimedia applications, Quality of Services (QoS) support in such infrastructure less networks have become essential. QoS routi…
▽ More
Mobile Ad-hoc Network (MANET) is the self organizing collection of mobile nodes. The communication in MANET is done via a wireless media. Ad hoc wireless networks have massive commercial and military potential because of their mobility support. Due to demanding real time multimedia applications, Quality of Services (QoS) support in such infrastructure less networks have become essential. QoS routing in mobile Ad-Hoc networks is challenging due to rapid change in network topology. Consequently, the available state information for routing is inherently imprecise. QoS routing may suffer badly due to several factors including radio interference on available bandwidth, and inefficient flooding of information to the adjacent nodes. As a result the performance of the network degrades substantially. This paper aims at the solution for energy efficient QoS routing by best utilization of network resources such as energy and bandwidth. A comparative study shows that despite the overhead due to QoS management, this solution performs better than classical OLSR protocol in terms of QoS and efficient utilization of energy.
△ Less
Submitted 20 August, 2011;
originally announced August 2011.
-
A New Trusted and Collaborative Agent Based Approach for Ensuring Cloud Security
Authors:
Shantanu Pal,
Sunirmal Khatua,
Nabendu Chaki,
Sugata Sanyal
Abstract:
In order to determine the user's trust is a growing concern for ensuring privacy and security in a cloud computing environment. In cloud, user's data is stored in one or more remote server(s) which poses more security challenges for the system. One of the most important concerns is to protect user's sensitive information from other users and hackers that may cause data leakage in cloud storage. Ha…
▽ More
In order to determine the user's trust is a growing concern for ensuring privacy and security in a cloud computing environment. In cloud, user's data is stored in one or more remote server(s) which poses more security challenges for the system. One of the most important concerns is to protect user's sensitive information from other users and hackers that may cause data leakage in cloud storage. Having this security challenge in mind, this paper focuses on the development of a more secure cloud environment, to determine the trust of the service requesting authorities by using a novel VM (Virtual Machine) monitoring system. Moreover, this research aims towards proposing a new trusted and collaborative agent-based two-tier framework, titled WAY (Who Are You?), to protect cloud resources. The framework can be used to provide security in network, infrastructure, as well as data storage in a heterogeneous cloud platform. If the trust updating policy is based on network activities, then the framework can provide network security. Similarly, it provides storage security by monitoring unauthorized access activities by the Cloud Service Users (CSU). Infrastructure security can be provided by monitoring the use of privileged instructions within the isolated VMs. The uniqueness of the proposed security solution lies in the fact that it ensures security and privacy both at the service provider level as well as at the user level in a cloud environment.
△ Less
Submitted 20 August, 2011;
originally announced August 2011.
-
A New Cluster-based Wormhole Intrusion detection algorithm for Mobile Ad-Hoc Networks
Authors:
Debdutta Barman Roy,
Rituparna Chaki,
Nabendu Chaki
Abstract:
In multi-hop wireless systems, the need for cooperation among nodes to relay each other's packets exposes them to a wide range of security attacks. A particularly devastating attack is the wormhole attack, where a malicious node records control traffic at one location and tunnels it to another compromised node, possibly far away, which replays it locally. Routing security in ad hoc networks is oft…
▽ More
In multi-hop wireless systems, the need for cooperation among nodes to relay each other's packets exposes them to a wide range of security attacks. A particularly devastating attack is the wormhole attack, where a malicious node records control traffic at one location and tunnels it to another compromised node, possibly far away, which replays it locally. Routing security in ad hoc networks is often equated with strong and feasible node authentication and lightweight cryptography. Unfortunately, the wormhole attack can hardly be defeated by crypto graphical measures, as wormhole attackers do not create separate packets. They simply replay packets already existing on the network, which pass the cryptographic checks. Existing works on wormhole detection have often focused on detection using specialized hardware, such as directional antennas, etc. In this paper, we present a cluster based counter-measure for the wormhole attack, that alleviates these drawbacks and efficiently mitigates the wormhole attack in MANET. Simulation results on MATLab exhibit the effectiveness of the proposed algorithm in detecting wormhole attacks.
△ Less
Submitted 5 April, 2010;
originally announced April 2010.
