Matching Items (46)
Filtering by
- All Subjects: Computer Science
- Creators: Bansal, Ajay
Description
Image understanding has been playing an increasingly crucial role in vision applications. Sparse models form an important component in image understanding, since the statistics of natural images reveal the presence of sparse structure. Sparse methods lead to parsimonious models, in addition to being efficient for large scale learning. In sparse modeling, data is represented as a sparse linear combination of atoms from a "dictionary" matrix. This dissertation focuses on understanding different aspects of sparse learning, thereby enhancing the use of sparse methods by incorporating tools from machine learning. With the growing need to adapt models for large scale data, it is important to design dictionaries that can model the entire data space and not just the samples considered. By exploiting the relation of dictionary learning to 1-D subspace clustering, a multilevel dictionary learning algorithm is developed, and it is shown to outperform conventional sparse models in compressed recovery, and image denoising. Theoretical aspects of learning such as algorithmic stability and generalization are considered, and ensemble learning is incorporated for effective large scale learning. In addition to building strategies for efficiently implementing 1-D subspace clustering, a discriminative clustering approach is designed to estimate the unknown mixing process in blind source separation. By exploiting the non-linear relation between the image descriptors, and allowing the use of multiple features, sparse methods can be made more effective in recognition problems. The idea of multiple kernel sparse representations is developed, and algorithms for learning dictionaries in the feature space are presented. Using object recognition experiments on standard datasets it is shown that the proposed approaches outperform other sparse coding-based recognition frameworks. Furthermore, a segmentation technique based on multiple kernel sparse representations is developed, and successfully applied for automated brain tumor identification. Using sparse codes to define the relation between data samples can lead to a more robust graph embedding for unsupervised clustering. By performing discriminative embedding using sparse coding-based graphs, an algorithm for measuring the glomerular number in kidney MRI images is developed. Finally, approaches to build dictionaries for local sparse coding of image descriptors are presented, and applied to object recognition and image retrieval.
ContributorsJayaraman Thiagarajan, Jayaraman (Author) / Spanias, Andreas (Thesis advisor) / Frakes, David (Committee member) / Tepedelenlioğlu, Cihan (Committee member) / Turaga, Pavan (Committee member) / Arizona State University (Publisher)
Created2013
Description
The Internet is transforming its look, in a short span of time we have come very far from black and white web forms with plain buttons to responsive, colorful and appealing user interface elements. With the sudden rise in demand of web applications, developers are making full use of the power of HTML5, JavaScript and CSS3 to cater to their users on various platforms. There was never a need of classifying the ways in which these languages can be interconnected to each other as the size of the front end code base was relatively small and did not involve critical business logic. This thesis focuses on listing and defining all dependencies between HTML5, JavaScript and CSS3 that will help developers better understand the interconnections within these languages. We also explore the present techniques available to a developer to make his code free of dependency related defects. We build a prototype tool, HJCDepend, based on our model, which aims at helping developers discover and remove defects early in the development cycle.
ContributorsVasugupta (Author) / Gary, Kevin (Thesis advisor) / Lindquist, Timothy (Committee member) / Bansal, Ajay (Committee member) / Arizona State University (Publisher)
Created2014
Description
The processing of large volumes of RDF data require an efficient storage and query processing engine that can scale well with the volume of data. The initial attempts to address this issue focused on optimizing native RDF stores as well as conventional relational databases management systems. But as the volume of RDF data grew to exponential proportions, the limitations of these systems became apparent and researchers began to focus on using big data analysis tools, most notably Hadoop, to process RDF data. Various studies and benchmarks that evaluate these tools for RDF data processing have been published. In the past two and half years, however, heavy users of big data systems, like Facebook, noted limitations with the query performance of these big data systems and began to develop new distributed query engines for big data that do not rely on map-reduce. Facebook's Presto is one such example.
This thesis deals with evaluating the performance of Presto in processing big RDF data against Apache Hive. A comparative analysis was also conducted against 4store, a native RDF store. To evaluate the performance Presto for big RDF data processing, a map-reduce program and a compiler, based on Flex and Bison, were implemented. The map-reduce program loads RDF data into HDFS while the compiler translates SPARQL queries into a subset of SQL that Presto (and Hive) can understand. The evaluation was done on four and eight node Linux clusters installed on Microsoft Windows Azure platform with RDF datasets of size 10, 20, and 30 million triples. The results of the experiment show that Presto has a much higher performance than Hive can be used to process big RDF data. The thesis also proposes an architecture based on Presto, Presto-RDF, that can be used to process big RDF data.
This thesis deals with evaluating the performance of Presto in processing big RDF data against Apache Hive. A comparative analysis was also conducted against 4store, a native RDF store. To evaluate the performance Presto for big RDF data processing, a map-reduce program and a compiler, based on Flex and Bison, were implemented. The map-reduce program loads RDF data into HDFS while the compiler translates SPARQL queries into a subset of SQL that Presto (and Hive) can understand. The evaluation was done on four and eight node Linux clusters installed on Microsoft Windows Azure platform with RDF datasets of size 10, 20, and 30 million triples. The results of the experiment show that Presto has a much higher performance than Hive can be used to process big RDF data. The thesis also proposes an architecture based on Presto, Presto-RDF, that can be used to process big RDF data.
ContributorsMammo, Mulugeta (Author) / Bansal, Srividya (Thesis advisor) / Bansal, Ajay (Committee member) / Lindquist, Timothy (Committee member) / Arizona State University (Publisher)
Created2014
Description
Gathering and managing software requirements, known as Requirement Engineering (RE), is a significant and basic step during the Software Development Life Cycle (SDLC). Any error or defect during the RE step will propagate to further steps of SDLC and resolving it will be more costly than any defect in other steps. In order to produce better quality software, the requirements have to be free of any defects. Verification and Validation (V&V;) of requirements are performed to improve their quality, by performing the V&V; process on the Software Requirement Specification (SRS) document. V&V; of the software requirements focused to a specific domain helps in improving quality. A large database of software requirements from software projects of different domains is created. Software requirements from commercial applications are focus of this project; other domains embedded, mobile, E-commerce, etc. can be the focus of future efforts. The V&V; is done to inspect the requirements and improve the quality. Inspections are done to detect defects in the requirements and three approaches for inspection of software requirements are discussed; ad-hoc techniques, checklists, and scenario-based techniques. A more systematic domain-specific technique is presented for performing V&V; of requirements.
ContributorsChughtai, Rehman (Author) / Ghazarian, Arbi (Thesis advisor) / Bansal, Ajay (Committee member) / Millard, Bruce (Committee member) / Arizona State University (Publisher)
Created2012
Description
Graph theory is a critical component of computer science and software engineering, with algorithms concerning graph traversal and comprehension powering much of the largest problems in both industry and research. Engineers and researchers often have an accurate view of their target graph, however they struggle to implement a correct, and efficient, search over that graph.
To facilitate rapid, correct, efficient, and intuitive development of graph based solutions we propose a new programming language construct - the search statement. Given a supra-root node, a procedure which determines the children of a given parent node, and optional definitions of the fail-fast acceptance or rejection of a solution, the search statement can conduct a search over any graph or network. Structurally, this statement is modelled after the common switch statement and is put into a largely imperative/procedural context to allow for immediate and intuitive development by most programmers. The Go programming language has been used as a foundation and proof-of-concept of the search statement. A Go compiler is provided which implements this construct.
To facilitate rapid, correct, efficient, and intuitive development of graph based solutions we propose a new programming language construct - the search statement. Given a supra-root node, a procedure which determines the children of a given parent node, and optional definitions of the fail-fast acceptance or rejection of a solution, the search statement can conduct a search over any graph or network. Structurally, this statement is modelled after the common switch statement and is put into a largely imperative/procedural context to allow for immediate and intuitive development by most programmers. The Go programming language has been used as a foundation and proof-of-concept of the search statement. A Go compiler is provided which implements this construct.
ContributorsHenderson, Christopher (Author) / Bansal, Ajay (Thesis advisor) / Lindquist, Timothy (Committee member) / Acuna, Ruben (Committee member) / Arizona State University (Publisher)
Created2018
Description
Modern machine learning systems leverage data and features from multiple modalities to gain more predictive power. In most scenarios, the modalities are vastly different and the acquired data are heterogeneous in nature. Consequently, building highly effective fusion algorithms is at the core to achieve improved model robustness and inferencing performance. This dissertation focuses on the representation learning approaches as the fusion strategy. Specifically, the objective is to learn the shared latent representation which jointly exploit the structural information encoded in all modalities, such that a straightforward learning model can be adopted to obtain the prediction.
We first consider sensor fusion, a typical multimodal fusion problem critical to building a pervasive computing platform. A systematic fusion technique is described to support both multiple sensors and descriptors for activity recognition. Targeted to learn the optimal combination of kernels, Multiple Kernel Learning (MKL) algorithms have been successfully applied to numerous fusion problems in computer vision etc. Utilizing the MKL formulation, next we describe an auto-context algorithm for learning image context via the fusion with low-level descriptors. Furthermore, a principled fusion algorithm using deep learning to optimize kernel machines is developed. By bridging deep architectures with kernel optimization, this approach leverages the benefits of both paradigms and is applied to a wide variety of fusion problems.
In many real-world applications, the modalities exhibit highly specific data structures, such as time sequences and graphs, and consequently, special design of the learning architecture is needed. In order to improve the temporal modeling for multivariate sequences, we developed two architectures centered around attention models. A novel clinical time series analysis model is proposed for several critical problems in healthcare. Another model coupled with triplet ranking loss as metric learning framework is described to better solve speaker diarization. Compared to state-of-the-art recurrent networks, these attention-based multivariate analysis tools achieve improved performance while having a lower computational complexity. Finally, in order to perform community detection on multilayer graphs, a fusion algorithm is described to derive node embedding from word embedding techniques and also exploit the complementary relational information contained in each layer of the graph.
We first consider sensor fusion, a typical multimodal fusion problem critical to building a pervasive computing platform. A systematic fusion technique is described to support both multiple sensors and descriptors for activity recognition. Targeted to learn the optimal combination of kernels, Multiple Kernel Learning (MKL) algorithms have been successfully applied to numerous fusion problems in computer vision etc. Utilizing the MKL formulation, next we describe an auto-context algorithm for learning image context via the fusion with low-level descriptors. Furthermore, a principled fusion algorithm using deep learning to optimize kernel machines is developed. By bridging deep architectures with kernel optimization, this approach leverages the benefits of both paradigms and is applied to a wide variety of fusion problems.
In many real-world applications, the modalities exhibit highly specific data structures, such as time sequences and graphs, and consequently, special design of the learning architecture is needed. In order to improve the temporal modeling for multivariate sequences, we developed two architectures centered around attention models. A novel clinical time series analysis model is proposed for several critical problems in healthcare. Another model coupled with triplet ranking loss as metric learning framework is described to better solve speaker diarization. Compared to state-of-the-art recurrent networks, these attention-based multivariate analysis tools achieve improved performance while having a lower computational complexity. Finally, in order to perform community detection on multilayer graphs, a fusion algorithm is described to derive node embedding from word embedding techniques and also exploit the complementary relational information contained in each layer of the graph.
ContributorsSong, Huan (Author) / Spanias, Andreas (Thesis advisor) / Thiagarajan, Jayaraman (Committee member) / Berisha, Visar (Committee member) / Tepedelenlioğlu, Cihan (Committee member) / Arizona State University (Publisher)
Created2018
Description
UVLabel was created to enable radio astronomers to view and annotate their own data such that they could then expand their future research paths. It simplifies their data rendering process by providing a simple user interface to better access sections of their data. Furthermore, it provides an interface to track trends in their data through a labelling feature.
The tool was developed following the incremental development process in order to quickly create a functional and testable tool. The incremental process also allowed for feedback from radio astronomers to help guide the project's development.
UVLabel provides both a functional product, and a modifiable and scalable code base for radio astronomer developers. This enables astronomers studying various astronomical interferometric data labelling capabilities. The tool can then be used to improve their filtering methods, pursue machine learning solutions, and discover new trends. Finally, UVLabel will be open source to put customization, scalability, and adaptability in the hands of these researchers.
The tool was developed following the incremental development process in order to quickly create a functional and testable tool. The incremental process also allowed for feedback from radio astronomers to help guide the project's development.
UVLabel provides both a functional product, and a modifiable and scalable code base for radio astronomer developers. This enables astronomers studying various astronomical interferometric data labelling capabilities. The tool can then be used to improve their filtering methods, pursue machine learning solutions, and discover new trends. Finally, UVLabel will be open source to put customization, scalability, and adaptability in the hands of these researchers.
ContributorsLa Place, Cecilia (Author) / Bansal, Ajay (Thesis advisor) / Jacobs, Daniel (Thesis advisor) / Acuna, Ruben (Committee member) / Arizona State University (Publisher)
Created2019
Description
Capturing the information in an image into a natural language sentence is
considered a difficult problem to be solved by computers. Image captioning involves not just detecting objects from images but understanding the interactions between the objects to be translated into relevant captions. So, expertise in the fields of computer vision paired with natural language processing are supposed to be crucial for this purpose. The sequence to sequence modelling strategy of deep neural networks is the traditional approach to generate a sequential list of words which are combined to represent the image. But these models suffer from the problem of high variance by not being able to generalize well on the training data.
The main focus of this thesis is to reduce the variance factor which will help in generating better captions. To achieve this, Ensemble Learning techniques have been explored, which have the reputation of solving the high variance problem that occurs in machine learning algorithms. Three different ensemble techniques namely, k-fold ensemble, bootstrap aggregation ensemble and boosting ensemble have been evaluated in this thesis. For each of these techniques, three output combination approaches have been analyzed. Extensive experiments have been conducted on the Flickr8k dataset which has a collection of 8000 images and 5 different captions for every image. The bleu score performance metric, which is considered to be the standard for evaluating natural language processing (NLP) problems, is used to evaluate the predictions. Based on this metric, the analysis shows that ensemble learning performs significantly better and generates more meaningful captions compared to any of the individual models used.
considered a difficult problem to be solved by computers. Image captioning involves not just detecting objects from images but understanding the interactions between the objects to be translated into relevant captions. So, expertise in the fields of computer vision paired with natural language processing are supposed to be crucial for this purpose. The sequence to sequence modelling strategy of deep neural networks is the traditional approach to generate a sequential list of words which are combined to represent the image. But these models suffer from the problem of high variance by not being able to generalize well on the training data.
The main focus of this thesis is to reduce the variance factor which will help in generating better captions. To achieve this, Ensemble Learning techniques have been explored, which have the reputation of solving the high variance problem that occurs in machine learning algorithms. Three different ensemble techniques namely, k-fold ensemble, bootstrap aggregation ensemble and boosting ensemble have been evaluated in this thesis. For each of these techniques, three output combination approaches have been analyzed. Extensive experiments have been conducted on the Flickr8k dataset which has a collection of 8000 images and 5 different captions for every image. The bleu score performance metric, which is considered to be the standard for evaluating natural language processing (NLP) problems, is used to evaluate the predictions. Based on this metric, the analysis shows that ensemble learning performs significantly better and generates more meaningful captions compared to any of the individual models used.
ContributorsKatpally, Harshitha (Author) / Bansal, Ajay (Thesis advisor) / Acuna, Ruben (Committee member) / Gonzalez-Sanchez, Javier (Committee member) / Arizona State University (Publisher)
Created2019
Description
Since the advent of the internet and even more after social media platforms, the explosive growth of textual data and its availability has made analysis a tedious task. Information extraction systems are available but are generally too specific and often only extract certain kinds of information they deem necessary and extraction worthy. Using data visualization theory and fast, interactive querying methods, leaving out information might not really be necessary. This thesis explores textual data visualization techniques, intuitive querying, and a novel approach to all-purpose textual information extraction to encode large text corpus to improve human understanding of the information present in textual data.
This thesis presents a modified traversal algorithm on dependency parse output of text to extract all subject predicate object pairs from text while ensuring that no information is missed out. To support full scale, all-purpose information extraction from large text corpuses, a data preprocessing pipeline is recommended to be used before the extraction is run. The output format is designed specifically to fit on a node-edge-node model and form the building blocks of a network which makes understanding of the text and querying of information from corpus quick and intuitive. It attempts to reduce reading time and enhancing understanding of the text using interactive graph and timeline.
This thesis presents a modified traversal algorithm on dependency parse output of text to extract all subject predicate object pairs from text while ensuring that no information is missed out. To support full scale, all-purpose information extraction from large text corpuses, a data preprocessing pipeline is recommended to be used before the extraction is run. The output format is designed specifically to fit on a node-edge-node model and form the building blocks of a network which makes understanding of the text and querying of information from corpus quick and intuitive. It attempts to reduce reading time and enhancing understanding of the text using interactive graph and timeline.
ContributorsHashmi, Syed Usama (Author) / Bansal, Ajay (Thesis advisor) / Bansal, Srividya (Committee member) / Gonzalez Sanchez, Javier (Committee member) / Arizona State University (Publisher)
Created2018
Description
Multiple-input multiple-output systems have gained focus in the last decade due to the benefits they provide in enhancing the quality of communications. On the other hand, full-duplex communication has attracted remarkable attention due to its ability to improve the spectral efficiency compared to the existing half-duplex systems. Using full-duplex communications on MIMO co-operative networks can provide us solutions that can completely outperform existing systems with simultaneous transmission and reception at high data rates.
This thesis considers a full-duplex MIMO relay which amplifies and forwards the received signals, between a source and a destination that do not a have line of sight. Full-duplex mode raises the problem of self-interference. Though all the links in the system undergo frequency flat fading, the end-to-end effective channel is frequency selective. This is due to the imperfect cancellation of the self-interference at the relay and this residual self-interference acts as intersymbol interference at the destination which is treated by equalization. This also leads to complications in form of recursive equations to determine the input-output relationship of the system. This also leads to complications in the form of recursive equations to determine the input-output relationship of the system.
To overcome this, a signal flow graph approach using Mason's gain formula is proposed, where the effective channel is analyzed with keen notice to every loop and path the signal traverses. This gives a clear understanding and awareness about the orders of the polynomials involved in the transfer function, from which desired conclusions can be drawn. But the complexity of Mason's gain formula increases with the number of antennas at relay which can be overcome by the proposed linear algebraic method. Input-output relationship derived using simple concepts of linear algebra can be generalized to any number of antennas and the computation complexity is comparatively very low.
For a full-duplex amplify-and-forward MIMO relay system, assuming equalization at the destination, new mechanisms have been implemented at the relay that can compensate the effect of residual self-interference namely equal-gain transmission and antenna selection. Though equal-gain transmission does not perform better than the maximal ratio transmission, a trade-off can be made between performance and implementation complexity. Using the proposed antenna selection strategy, one pair of transmit-receive antennas at the relay is selected based on four selection criteria discussed. Outage probability analysis is performed for all the strategies presented and detailed comparison has been established. Considering minimum mean-squared error decision feedback equalizer at the destination, a bound on the outage probability has been obtained for the antenna selection case and is used for comparisons. A cross-over point is observed while comparing the outage probabilities of equal-gain transmission and antenna selection techniques, as the signal-to-noise ratio increases and from that point antenna selection outperforms equal-gain transmission and this is explained by the fact of reduced residual self-interference in antenna selection method.
This thesis considers a full-duplex MIMO relay which amplifies and forwards the received signals, between a source and a destination that do not a have line of sight. Full-duplex mode raises the problem of self-interference. Though all the links in the system undergo frequency flat fading, the end-to-end effective channel is frequency selective. This is due to the imperfect cancellation of the self-interference at the relay and this residual self-interference acts as intersymbol interference at the destination which is treated by equalization. This also leads to complications in form of recursive equations to determine the input-output relationship of the system. This also leads to complications in the form of recursive equations to determine the input-output relationship of the system.
To overcome this, a signal flow graph approach using Mason's gain formula is proposed, where the effective channel is analyzed with keen notice to every loop and path the signal traverses. This gives a clear understanding and awareness about the orders of the polynomials involved in the transfer function, from which desired conclusions can be drawn. But the complexity of Mason's gain formula increases with the number of antennas at relay which can be overcome by the proposed linear algebraic method. Input-output relationship derived using simple concepts of linear algebra can be generalized to any number of antennas and the computation complexity is comparatively very low.
For a full-duplex amplify-and-forward MIMO relay system, assuming equalization at the destination, new mechanisms have been implemented at the relay that can compensate the effect of residual self-interference namely equal-gain transmission and antenna selection. Though equal-gain transmission does not perform better than the maximal ratio transmission, a trade-off can be made between performance and implementation complexity. Using the proposed antenna selection strategy, one pair of transmit-receive antennas at the relay is selected based on four selection criteria discussed. Outage probability analysis is performed for all the strategies presented and detailed comparison has been established. Considering minimum mean-squared error decision feedback equalizer at the destination, a bound on the outage probability has been obtained for the antenna selection case and is used for comparisons. A cross-over point is observed while comparing the outage probabilities of equal-gain transmission and antenna selection techniques, as the signal-to-noise ratio increases and from that point antenna selection outperforms equal-gain transmission and this is explained by the fact of reduced residual self-interference in antenna selection method.
ContributorsJonnalagadda, Geeta Sankar Kalyan (Author) / Tepedelenlioğlu, Cihan (Thesis advisor) / Bliss, Daniel (Committee member) / Kosut, Oliver (Committee member) / Arizona State University (Publisher)
Created2018