Papers by Vijay Agneeswaran
Deep Learning and the Human Brain: Architectural and Learning Differences
Academia letters, Aug 13, 2021
arXiv (Cornell University), Jul 27, 2022
Missing values, widely called as sparsity in literature, is a common characteristic of many real-... more Missing values, widely called as sparsity in literature, is a common characteristic of many real-world datasets. Many imputation methods have been proposed to address this problem of data incompleteness or sparsity. However, the accuracy of a data imputation method for a given feature or a set of features in a dataset is highly dependent on the distribution of the feature values and its correlation with other features. Another problem that plagues industry deployments of machine learning (ML) solutions is concept drift detection, which becomes more challenging in the presence of missing values. Although data imputation and concept drift detection have been studied extensively, little work has attempted a combined study of the two phenomena, i.e., concept drift detection in the presence of sparsity. In this work, we carry out a systematic study of the following: (i) different patterns of missing values, (ii) various statistical and ML based data imputation methods for different kinds of sparsity, (iii) several concept drift detection methods, (iv) practical analysis of the various drift detection metrics, (v) selecting the best concept drift detector given a dataset with missing values based on the different metrics. We first analyze it on synthetic data and publicly available datasets, and finally extend the findings to our deployed solution of automated change risk assessment system. One of the major findings from our empirical study is the absence of supremacy of any one concept drift detection method across all the relevant metrics. Therefore, we adopt a majority voting based ensemble of concept drift detectors for abrupt and gradual concept drifts. Our experiments show optimal or near optimal performance can be achieved for this ensemble method across all the metrics.
Our data scientists are adept in using machine learning algorithms and building model out of it, ... more Our data scientists are adept in using machine learning algorithms and building model out of it, and they are at ease with their local machine to do them. But, when it comes to building the same model from the platform, they find it slightly challenging and need assistance from the platform team. Based on the survey results, the major challenge was platform complexity, but it is hard to deduce actionable items or accurate details to make the system simple. The complexity feedback was very generic, so we decided to break it down into two logical challenges: Education & Training and Simplicity-of-Platform. We have developed a system to find these two challenges in our platform, which we call an Analyzer. In this paper, we explain how it was built and it's impact on the evolution of our machine learning platform. Our work aims to address these challenges and provide guidelines on how to empower machine learning platform team to know the data scientist's bottleneck in building model.
Proceedings of the 17th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, 2022
Scene Text Recognition (STR) enables processing and understanding of the text in the wild. Howeve... more Scene Text Recognition (STR) enables processing and understanding of the text in the wild. However, roadblocks like natural degradation, blur, and uneven lighting in the captured images result in poor accuracy during detection and recognition. Previous approaches have introduced Super-Resolution (SR) as a processing step between detection and recognition; however, post enhancement, there is a significant drop in the quality of the reconstructed text in the image. This drop is especially significant in the healthcare domain because any loss in accuracy can be detrimental. This paper will quantitatively show the drop in quality of the text in an image from the existing SR techniques across multiple optimization-based and GAN-based models. We propose a new loss function for training and an improved deep neural network architecture to address these shortcomings and recover text with sharp boundaries in the SR images. We also show that the Peak Signal-to-Noise Ratio (PSNR) and the Structural Similarity Index Measure (SSIM) scores are not effective metrics for identifying the quality of the text in an SR image. Extensive experiments show that our model achieves better accuracy and visual improvements against state-of-the-art methods in terms of text recognition accuracy. We plan to add our module on SR in the near future to our already deployed solution for text extraction from product images for our company.
Extracting texts of various shapes, sizes, and orientations from images containing multiple objec... more Extracting texts of various shapes, sizes, and orientations from images containing multiple objects is an important problem in many contexts, especially, in connection to E-commerce. In the context of the scale at which Walmart operates, the text from an image can be a richer and more accurate source of data than human inputs and can be used in several applications such as Attribute Extraction, Offensive Text Classification, Product Matching among others. The motivation of this particular work has come from different business requirements such as flagging products whose images contain words that are non-compliant with organizational policies and building an efficient automated system to identify similar products by comparing the information contained in their respective product images and many others. Existing methods fail to address domain specific challenges like high entropy, different orientations, and small texts in product images adequately. In this work, we provide a solution that not only addresses these challenges but is also proven to work at a million image scale for various retail business units within Walmart. Extensive experimentation revealed that our proposed solution has been able to save around 30% computational cost in both the training and the inference stages. CCS CONCEPTS • Computing methodologies → Object detection; Object recognition; • Software and its engineering → Software design techniques.
This paper makes a step in identifying the state of the art in semantic P2P systems. On one hand,... more This paper makes a step in identifying the state of the art in semantic P2P systems. On one hand, lot of research in the P2P systems community has focused on fault-tolerance and scalability, resulting in numberous algorithms, systems such as Chord, Pastry and P-Grid. These systems, however, have no notion of semantics and consequently, have difficulty in knowledge sharing. On the other hand, research in the semantic web community have focused on knowledge sharing among different nodes with possibly different schemas. These have tended to use centralized repositories. The obvious benefits of combining P2P and semantic systems would be to have large scale collection of structured data. Several recent efforts have focused on this combination. However, there have been no attempt to have these efforts grouped in one place for easy assimilation and for finding interesting future directions; this paper fills the gap.
A Peer-to-Peer Framework for Collaborative Data Sharing over the Internet
Collaborative Computing, 2006
A complex system inspired theoretical formalism for data management in Peer-to-Peer grids
Proceedings, Apr 1, 2008
... Vijay Srinivas Agneeswaran Distributed Information Systems Lab (LSIR), School of Information ... more ... Vijay Srinivas Agneeswaran Distributed Information Systems Lab (LSIR), School of Information and Communication Sciences Swiss Federal Institute of Technology, Lausanne(EPFL), Switzerland. ... 5We are grateful to Dr. Rajkumar Buyya for providing us with these nodes. ...
arXiv (Cornell University), Apr 13, 2023
Vision transformers have been applied successfully for image recognition tasks. There have been e... more Vision transformers have been applied successfully for image recognition tasks. There have been either multi-headed self-attention based (ViT [14], DeIT, [53]) similar to the original work in textual models or more recently based on spectral layers (Fnet[29], GFNet[47], AFNO[17]). We hypothesize that both spectral and multi-headed attention plays a major role. We investigate this hypothesis through this work and observe that indeed combining spectral and multiheaded attention layers provides a better transformer architecture. We thus propose the novel Spectformer architecture for transformers that combines spectral and multi-headed attention layers. We believe that the resulting representation allows the transformer to capture the feature representation appropriately and it yields improved performance over other transformer representations. For instance, it improves the top-1 accuracy by 2% on ImageNet compared to both GFNet-H and LiT. SpectFormer-S reaches 84.25% top-1 accuracy on ImageNet-1K (state of the art for small version). Further, Spectformer-L achieves 85.7% that is the state of the art for the comparable base version of the transformers. We further ensure that we obtain reasonable results in other scenarios such as transfer learning on standard datasets such as CIFAR-10, CIFAR-100, Oxford-IIIT-flower, and Standford Car datasets. We then investigate its use in downstream tasks such of object detection and instance segmentation on MS-COCO dataset and observe that Spectformer shows consistent performance that is comparable to the best backbones and can be further optimized and improved. Hence, we believe that combined spectral and attention layers are what are needed for vision transformers. The project page is available at this webpage.https: //badripatro.github.io/SpectFormers/.
arXiv (Cornell University), Feb 16, 2023
Transformers are widely used for solving tasks in natural language processing, computer vision, s... more Transformers are widely used for solving tasks in natural language processing, computer vision, speech, and music domains. In this paper, we talk about the efficiency of transformers in terms of memory (the number of parameters), computation cost (number of floating points operations), and performance of models, including accuracy, robustness of the model, and fair & bias-free features. We mainly discuss the vision transformer for the image classification task. Our contribution is to introduce an efficient 360 framework, which includes various aspects of the vision transformer, to make it more efficient for industrial applications. By considering those applications, we categorize them into multiple dimensions such as privacy, robustness, transparency, fairness, inclusiveness, continual learning, probabilistic models, approximation, computational complexity, and spectral complexity. We compare various vision transformer models based on their performance, the number of parameters, and the number of floating point operations (FLOPs) on multiple datasets.
Object migration in CORBA
Abstract: The power and flexibility of a distributed object system increases many fold if it allo... more Abstract: The power and flexibility of a distributed object system increases many fold if it allows objects to migrate across machines. CORBA, which is emerging as a standard for distributed object computing, currently lacks a proper mechanism for object migration. In ...
arXiv (Cornell University), Jul 27, 2022
Missing values, widely called as sparsity in literature, is a common characteristic of many real-... more Missing values, widely called as sparsity in literature, is a common characteristic of many real-world datasets. Many imputation methods have been proposed to address this problem of data incompleteness or sparsity. However, the accuracy of a data imputation method for a given feature or a set of features in a dataset is highly dependent on the distribution of the feature values and its correlation with other features. Another problem that plagues industry deployments of machine learning (ML) solutions is concept drift detection, which becomes more challenging in the presence of missing values. Although data imputation and concept drift detection have been studied extensively, little work has attempted a combined study of the two phenomena, i.e., concept drift detection in the presence of sparsity. In this work, we carry out a systematic study of the following: (i) different patterns of missing values, (ii) various statistical and ML based data imputation methods for different kinds of sparsity, (iii) several concept drift detection methods, (iv) practical analysis of the various drift detection metrics, (v) selecting the best concept drift detector given a dataset with missing values based on the different metrics. We first analyze it on synthetic data and publicly available datasets, and finally extend the findings to our deployed solution of automated change risk assessment system. One of the major findings from our empirical study is the absence of supremacy of any one concept drift detection method across all the relevant metrics. Therefore, we adopt a majority voting based ensemble of concept drift detectors for abrupt and gradual concept drifts. Our experiments show optimal or near optimal performance can be achieved for this ensemble method across all the metrics.
This paper makes a step in identifying the state of the art in semantic P2P systems. On one hand,... more This paper makes a step in identifying the state of the art in semantic P2P systems. On one hand, lot of research in the P2P systems community has focused on fault-tolerance and scalability, resulting in numberous algorithms, systems such as Chord, Pastry and P-Grid. These systems, however, have no notion of semantics and consequently, have difficulty in knowledge sharing. On the other hand, research in the semantic web community have focused on knowledge sharing among different nodes with possibly different schemas. These have tended to use centralized repositories. The obvious benefits of combining P2P and semantic systems would be to have large scale collection of structured data. Several recent efforts have focused on this combination. However, there have been no attempt to have these efforts grouped in one place for easy assimilation and for finding interesting future directions; this paper fills the gap.
SpringerReference
The current standard for Fault-Tolerance in the Common Object Request Broker Architecture (CORBA)... more The current standard for Fault-Tolerance in the Common Object Request Broker Architecture (CORBA) does not support network partitioning. However, distributed systems, and those deployed on wide area networks in particular, are susceptible to network partitions. The contribution of this paper is the description of the design and implementation of a CORBA fault-tolerance add-on for partionable environments. Our solution can be applied to an off-the-shelf Object Request Broker, without having access to the ORB's source code and with minimal changes to existing CORBA applications. The system distinguishes itself from existing solutions in the way different replication and reconciliation strategies can be implemented easily. Furthermore, we provide a novel replication and reconciliation protocol that increases the availability of systems, by allowing operations in all partitions, including nonmajority partitons to continue.
CCGrid 2011 Reviewers
computer.org
Sriram Aananthakrishnan Sherif Abdelwahed Vijay Srinivas Agneeswaran Muhammad Aleem Nawab Ali Ath... more Sriram Aananthakrishnan Sherif Abdelwahed Vijay Srinivas Agneeswaran Muhammad Aleem Nawab Ali Athanasios Antoniou Anne Auger Jim Basney Leonardo Bautista Josep Ll. Berral Laurent Bobelin Ryan Braithwaite John Bresnahan Rodrigo N. Calheiros Miguel Camelo Ghislain Charrier Qian Chen Wei-Fan Chiang Nitin Chiluka Pierre-Nicolas Clauss Xabriel J. Collazo-Mojica Minh Quan Dang Thomas De Ruiter Simon Delamare Javier Delgado Benjamin Depardon Marcos Dias De Assuncao James Dinan Mohammed El Mehdi Diouri Bruno Donassolo Abhishek ...
Data Management in Distributed Systems:A Scalability Taxonomy
Scalable Computing: Practice and Experience, 2007
Data management is a key aspect of any distributed system. This paper surveys data management tec... more Data management is a key aspect of any distributed system. This paper surveys data management techniques in various distributed systems, starting from Distributed Shared Memory (DSM) systems to Peer-to-Peer (P2P) systems. The central focus is on scalability, an important non-functional property of distributed systems. A scalability taxonomy of data management techniques is presented. Detailed discussion of the evolution of data management techniques in the different categories as well as the state of the art is provided. As a result, several open issues are inferred including use of P2P techniques in data grids and distributed mobile systems and the use of optimal data placement heuristics from Content Distribution Networks (CDNs) for P2P grids.
Proceedings of the 17th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications
Scene Text Recognition (STR) enables processing and understanding of the text in the wild. Howeve... more Scene Text Recognition (STR) enables processing and understanding of the text in the wild. However, roadblocks like natural degradation, blur, and uneven lighting in the captured images result in poor accuracy during detection and recognition. Previous approaches have introduced Super-Resolution (SR) as a processing step between detection and recognition; however, post enhancement, there is a significant drop in the quality of the reconstructed text in the image. This drop is especially significant in the healthcare domain because any loss in accuracy can be detrimental. This paper will quantitatively show the drop in quality of the text in an image from the existing SR techniques across multiple optimization-based and GAN-based models. We propose a new loss function for training and an improved deep neural network architecture to address these shortcomings and recover text with sharp boundaries in the SR images. We also show that the Peak Signal-to-Noise Ratio (PSNR) and the Structural Similarity Index Measure (SSIM) scores are not effective metrics for identifying the quality of the text in an SR image. Extensive experiments show that our model achieves better accuracy and visual improvements against state-of-the-art methods in terms of text recognition accuracy. We plan to add our module on SR in the near future to our already deployed solution for text extraction from product images for our company.
IEEE Distributed Systems Online, 2006
We're part of a group that's realizing Vishwa (http://dos.iitm.ac.in/Vishwa), a peer-topeer middl... more We're part of a group that's realizing Vishwa (http://dos.iitm.ac.in/Vishwa), a peer-topeer middleware architecture for developing grid applications. Two of us had a research brainstorming session with our advisor (the third author) during tea one fine day. We present parts of the session here to get across the key issues in building services for largescale distributed systems. VIJAY : Researchers are producing large amounts of scientific data for instance, see the Grid Physics Network Project (http://www.griphyn.org). Distributed computations on this data must be scheduled, and this data must be available to a large number of scientists. So, there's a need to replicate the data at appropriate locations to handle node and network failures and minimize computation time, bandwidth, or both. We wish to develop a platform that could serve as a substrate for building the replication service required for large P2P data grids. ADVISOR : This platform sounds interesting, but what are its key requirements? VENKAT : Scalability. The platform must scale up to a large number (possibly millions) of nodes and data units (objects) in the system. The platform must also be scalable geographically; that is, it must work well over the Internet.
A Model-Agnostic Framework to Correct Label-Bias in Training Data Using a Sample of Trusted Data
Advances in Intelligent Systems and Computing, 2021
Uploads
Papers by Vijay Agneeswaran