Matching Items (361)
Filtering by
- All Subjects: Machine Learning
- Member of: Theses and Dissertations
Description
Significance of real-world knowledge for Natural Language Understanding(NLU) is well-known for decades. With advancements in technology, challenging tasks like question-answering, text-summarizing, and machine translation are made possible with continuous efforts in the field of Natural Language Processing(NLP). Yet, knowledge integration to answer common sense questions is still a daunting task. Logical reasoning has been a resort for many of the problems in NLP and has achieved considerable results in the field, but it is difficult to resolve the ambiguities in a natural language. Co-reference resolution is one of the problems where ambiguity arises due to the semantics of the sentence. Another such problem is the cause and result statements which require causal commonsense reasoning to resolve the ambiguity. Modeling these type of problems is not a simple task with rules or logic. State-of-the-art systems addressing these problems use a trained neural network model, which claims to have overall knowledge from a huge trained corpus. These systems answer the questions by using the knowledge embedded in their trained language model. Although the language models embed the knowledge from the data, they use occurrences of words and frequency of co-existing words to solve the prevailing ambiguity. This limits the performance of language models to solve the problems in common-sense reasoning task as it generalizes the concept rather than trying to answer the problem specific to its context. For example, "The painting in Mark's living room shows an oak tree. It is to the right of a house", is a co-reference resolution problem which requires knowledge. Language models can resolve whether "it" refers to "painting" or "tree", since "house" and "tree" are two common co-occurring words so the models can resolve "tree" to be the co-reference. On the other hand, "The large ball crashed right through the table. Because it was made of Styrofoam ." to resolve for "it" which can be either "table" or "ball", is difficult for a language model as it requires more information about the problem.
In this work, I have built an end-to-end framework, which uses the automatically extracted knowledge based on the problem. This knowledge is augmented with the language models using an explicit reasoning module to resolve the ambiguity. This system is built to improve the accuracy of the language models based approaches for commonsense reasoning. This system has proved to achieve the state of the art accuracy on the Winograd Schema Challenge.
In this work, I have built an end-to-end framework, which uses the automatically extracted knowledge based on the problem. This knowledge is augmented with the language models using an explicit reasoning module to resolve the ambiguity. This system is built to improve the accuracy of the language models based approaches for commonsense reasoning. This system has proved to achieve the state of the art accuracy on the Winograd Schema Challenge.
ContributorsPrakash, Ashok (Author) / Baral, Chitta (Thesis advisor) / Devarakonda, Murthy (Committee member) / Anwar, Saadat (Committee member) / Arizona State University (Publisher)
Created2019
Description
Live streaming has risen to significant popularity in the recent past and largely this live streaming is a feature of existing social networks like Facebook, Instagram, and Snapchat. However, there does exist at least one social network entirely devoted to live streaming, and specifically the live streaming of video games, Twitch. This social network is unique for a number of reasons, not least because of its hyper-focus on live content and this uniqueness has challenges for social media researchers.
Despite this uniqueness, almost no scientific work has been performed on this public social network. Thus, it is unclear what user interaction features present on other social networks exist on Twitch. Investigating the interactions between users and identifying which, if any, of the common user behaviors on social network exist on Twitch is an important step in understanding how Twitch fits in to the social media ecosystem. For example, there are users that have large followings on Twitch and amass a large number of viewers, but do those users exert influence over the behavior of other user the way that popular users on Twitter do?
This task, however, will not be trivial. The same hyper-focus on live content that makes Twitch unique in the social network space invalidates many of the traditional approaches to social network analysis. Thus, new algorithms and techniques must be developed in order to tap this data source. In this thesis, a novel algorithm for finding games whose releases have made a significant impact on the network is described as well as a novel algorithm for detecting and identifying influential players of games. In addition, the Twitch network is described in detail along with the data that was collected in order to power the two previously described algorithms.
Despite this uniqueness, almost no scientific work has been performed on this public social network. Thus, it is unclear what user interaction features present on other social networks exist on Twitch. Investigating the interactions between users and identifying which, if any, of the common user behaviors on social network exist on Twitch is an important step in understanding how Twitch fits in to the social media ecosystem. For example, there are users that have large followings on Twitch and amass a large number of viewers, but do those users exert influence over the behavior of other user the way that popular users on Twitter do?
This task, however, will not be trivial. The same hyper-focus on live content that makes Twitch unique in the social network space invalidates many of the traditional approaches to social network analysis. Thus, new algorithms and techniques must be developed in order to tap this data source. In this thesis, a novel algorithm for finding games whose releases have made a significant impact on the network is described as well as a novel algorithm for detecting and identifying influential players of games. In addition, the Twitch network is described in detail along with the data that was collected in order to power the two previously described algorithms.
ContributorsJones, Isaac (Author) / Liu, Huan (Thesis advisor) / Maciejewski, Ross (Committee member) / Shakarian, Paulo (Committee member) / Agarwal, Nitin (Committee member) / Arizona State University (Publisher)
Created2019
Description
Over the last decade, deep neural networks also known as deep learning, combined with large databases and specialized hardware for computation, have made major strides in important areas such as computer vision, computational imaging and natural language processing. However, such frameworks currently suffer from some drawbacks. For example, it is generally not clear how the architectures are to be designed for different applications, or how the neural networks behave under different input perturbations and it is not easy to make the internal representations and parameters more interpretable. In this dissertation, I propose building constraints into feature maps, parameters and and design of algorithms involving neural networks for applications in low-level vision problems such as compressive imaging and multi-spectral image fusion, and high-level inference problems including activity and face recognition. Depending on the application, such constraints can be used to design architectures which are invariant/robust to certain nuisance factors, more efficient and, in some cases, more interpretable. Through extensive experiments on real-world datasets, I demonstrate these advantages of the proposed methods over conventional frameworks.
ContributorsLohit, Suhas Anand (Author) / Turaga, Pavan (Thesis advisor) / Spanias, Andreas (Committee member) / Li, Baoxin (Committee member) / Jayasuriya, Suren (Committee member) / Arizona State University (Publisher)
Created2019
Description
Longitudinal recursive partitioning (LRP) is a tree-based method for longitudinal data. It takes a sample of individuals that were each measured repeatedly across time, and it splits them based on a set of covariates such that individuals with similar trajectories become grouped together into nodes. LRP does this by fitting a mixed-effects model to each node every time that it becomes partitioned and extracting the deviance, which is the measure of node purity. LRP is implemented using the classification and regression tree algorithm, which suffers from a variable selection bias and does not guarantee reaching a global optimum. Additionally, fitting mixed-effects models to each potential split only to extract the deviance and discard the rest of the information is a computationally intensive procedure. Therefore, in this dissertation, I address the high computational demand, variable selection bias, and local optimum solution. I propose three approximation methods that reduce the computational demand of LRP, and at the same time, allow for a straightforward extension to recursive partitioning algorithms that do not have a variable selection bias and can reach the global optimum solution. In the three proposed approximations, a mixed-effects model is fit to the full data, and the growth curve coefficients for each individual are extracted. Then, (1) a principal component analysis is fit to the set of coefficients and the principal component score is extracted for each individual, (2) a one-factor model is fit to the coefficients and the factor score is extracted, or (3) the coefficients are summed. The three methods result in each individual having a single score that represents the growth curve trajectory. Therefore, now that the outcome is a single score for each individual, any tree-based method may be used for partitioning the data and group the individuals together. Once the individuals are assigned to their final nodes, a mixed-effects model is fit to each terminal node with the individuals belonging to it.
I conduct a simulation study, where I show that the approximation methods achieve the goals proposed while maintaining a similar level of out-of-sample prediction accuracy as LRP. I then illustrate and compare the methods using an applied data.
I conduct a simulation study, where I show that the approximation methods achieve the goals proposed while maintaining a similar level of out-of-sample prediction accuracy as LRP. I then illustrate and compare the methods using an applied data.
ContributorsStegmann, Gabriela (Author) / Grimm, Kevin (Thesis advisor) / Edwards, Michael (Committee member) / MacKinnon, David (Committee member) / McNeish, Daniel (Committee member) / Arizona State University (Publisher)
Created2019
Description
Disentangling latent spaces is an important research direction in the interpretability of unsupervised machine learning. Several recent works using deep learning are very effective at producing disentangled representations. However, in the unsupervised setting, there is no way to pre-specify which part of the latent space captures specific factors of variations. While this is generally a hard problem because of the non-existence of analytical expressions to capture these variations, there are certain factors like geometric
transforms that can be expressed analytically. Furthermore, in existing frameworks, the disentangled values are also not interpretable. The focus of this work is to disentangle these geometric factors of variations (which turn out to be nuisance factors for many applications) from the semantic content of the signal in an interpretable manner which in turn makes the features more discriminative. Experiments are designed to show the modularity of the approach with other disentangling strategies as well as on multiple one-dimensional (1D) and two-dimensional (2D) datasets, clearly indicating the efficacy of the proposed approach.
transforms that can be expressed analytically. Furthermore, in existing frameworks, the disentangled values are also not interpretable. The focus of this work is to disentangle these geometric factors of variations (which turn out to be nuisance factors for many applications) from the semantic content of the signal in an interpretable manner which in turn makes the features more discriminative. Experiments are designed to show the modularity of the approach with other disentangling strategies as well as on multiple one-dimensional (1D) and two-dimensional (2D) datasets, clearly indicating the efficacy of the proposed approach.
ContributorsKoneripalli Seetharam, Kaushik (Author) / Turaga, Pavan (Thesis advisor) / Papandreou-Suppappola, Antonia (Committee member) / Jayasuriya, Suren (Committee member) / Arizona State University (Publisher)
Created2019
Description
Humans perceive the environment using multiple modalities like vision, speech (language), touch, taste, and smell. The knowledge obtained from one modality usually complements the other. Learning through several modalities helps in constructing an accurate model of the environment. Most of the current vision and language models are modality-specific and, in many cases, extensively use deep-learning based attention mechanisms for learning powerful representations. This work discusses the role of attention in associating vision and language for generating shared representation. Language Image Transformer (LIT) is proposed for learning multi-modal representations of the environment. It uses a training objective based on Contrastive Predictive Coding (CPC) to maximize the Mutual Information (MI) between the visual and linguistic representations. It learns the relationship between the modalities using the proposed cross-modal attention layers. It is trained and evaluated using captioning datasets, MS COCO, and Conceptual Captions. The results and the analysis offers a perspective on the use of Mutual Information Maximisation (MIM) for generating generalizable representations across multiple modalities.
ContributorsRamakrishnan, Raghavendran (Author) / Panchanathan, Sethuraman (Thesis advisor) / Venkateswara, Hemanth Kumar (Thesis advisor) / McDaniel, Troy (Committee member) / Arizona State University (Publisher)
Created2020
Description
Question answering is a challenging problem and a long term goal of Artificial Intelligence. There are many approaches proposed to solve this problem, including end to end machine learning systems, Information Retrieval based approaches and Textual Entailment. Despite being popular, these methods find difficulty in solving problems that require multi level reasoning and combining independent pieces of knowledge, for example, a question like "What adaptation is necessary in intertidal ecosystems but not in reef ecosystems?'', requires the system to consider qualities, behaviour or features of an organism living in an intertidal ecosystem and compare with that of an organism in a reef ecosystem to find the answer. The proposed solution is to solve a genre of questions, which is questions based on "Adaptation, Variation and Behavior in Organisms", where there are various different independent sets of knowledge required for answering questions along with reasoning. This method is implemented using Answer Set Programming and Natural Language Inference (which is based on machine learning ) for finding which of the given options is more probable to be the answer by matching it with the knowledge base. To evaluate this approach, a dataset of questions and a knowledge base in the domain of "Adaptation, Variation and Behavior in Organisms" is created.
ContributorsBatni, Vaishnavi (Author) / Baral, Chitta (Thesis advisor) / Anwar, Saadat (Committee member) / Yang, Yezhou (Committee member) / Arizona State University (Publisher)
Created2019
Description
Driving is the coordinated operation of mind and body for movement of a vehicle, such as a car, or a bus. Driving, being considered an everyday activity for many people, still has an issue of safety. Driver distraction is becoming a critical safety problem. Speed, drunk driving as well as distracted driving are the three leading factors in the fatal car crashes. Distraction, which is defined as an excessive workload and limited attention, is the main paradigm that guides this research area. Driver behavior analysis can be used to address the distraction problem and provide an intelligent adaptive agent to work closely with the driver, fay beyond traditional algorithmic computational models. A variety of machine learning approaches has been proposed to estimate or predict drivers’ fatigue level using car data, driver status or a combination of them.
Three important features of intelligence and cognition are perception, attention and sensory memory. In this thesis, I focused on memory and attention as essential parts of highly intelligent systems. Without memory, systems will only show limited intelligence since their response would be exclusively based on spontaneous decision without considering the effect of previous events. I proposed a memory-based sequence to predict the driver behavior and distraction level using neural network. The work started with a large-scale experiment to collect data and make an artificial intelligence-friendly dataset. After that, the data was used to train a deep neural network to estimate the driver behavior. With a focus on memory by using Long Short Term Memory (LSTM) network to increase the level of intelligence in two dimensions: Forgiveness of minor glitches, and accumulation of anomalous behavior., I reduced the model error and computational expense by adding attention mechanism on the top of LSTM models. This system can be generalized to build and train highly intelligent agents in other domains.
Three important features of intelligence and cognition are perception, attention and sensory memory. In this thesis, I focused on memory and attention as essential parts of highly intelligent systems. Without memory, systems will only show limited intelligence since their response would be exclusively based on spontaneous decision without considering the effect of previous events. I proposed a memory-based sequence to predict the driver behavior and distraction level using neural network. The work started with a large-scale experiment to collect data and make an artificial intelligence-friendly dataset. After that, the data was used to train a deep neural network to estimate the driver behavior. With a focus on memory by using Long Short Term Memory (LSTM) network to increase the level of intelligence in two dimensions: Forgiveness of minor glitches, and accumulation of anomalous behavior., I reduced the model error and computational expense by adding attention mechanism on the top of LSTM models. This system can be generalized to build and train highly intelligent agents in other domains.
ContributorsMonjezi Kouchak, Shokoufeh (Author) / Gaffar, Ashraf (Thesis advisor) / Doupe, Adam (Committee member) / Ben Amor, Hani (Committee member) / Cheeks, Loretta (Committee member) / Arizona State University (Publisher)
Created2020
Description
Feature embeddings differ from raw features in the sense that the former obey certain properties like notion of similarity/dissimilarity in it's embedding space. word2vec is a preeminent example in this direction, where the similarity in the embedding space is measured in terms of the cosine similarity. Such language embedding models have seen numerous applications in both language and vision community as they capture the information in the modality (English language) efficiently. Inspired by these language models, this work focuses on learning embedding spaces for two visual computing tasks, 1. Image Hashing 2. Zero Shot Learning. The training set was used to learn embedding spaces over which similarity/dissimilarity is measured using several distance metrics like hamming / euclidean / cosine distances. While the above-mentioned language models learn generic word embeddings, in this work task specific embeddings were learnt which can be used for Image Retrieval and Classification separately.
Image Hashing is the task of mapping images to binary codes such that some notion of user-defined similarity is preserved. The first part of this work focuses on designing a new framework that uses the hash-tags associated with web images to learn the binary codes. Such codes can be used in several applications like Image Retrieval and Image Classification. Further, this framework requires no labelled data, leaving it very inexpensive. Results show that the proposed approach surpasses the state-of-art approaches by a significant margin.
Zero-shot classification is the task of classifying the test sample into a new class which was not seen during training. This is possible by establishing a relationship between the training and the testing classes using auxiliary information. In the second part of this thesis, a framework is designed that trains using the handcrafted attribute vectors and word vectors but doesn’t require the expensive attribute vectors during test time. More specifically, an intermediate space is learnt between the word vector space and the image feature space using the hand-crafted attribute vectors. Preliminary results on two zero-shot classification datasets show that this is a promising direction to explore.
Image Hashing is the task of mapping images to binary codes such that some notion of user-defined similarity is preserved. The first part of this work focuses on designing a new framework that uses the hash-tags associated with web images to learn the binary codes. Such codes can be used in several applications like Image Retrieval and Image Classification. Further, this framework requires no labelled data, leaving it very inexpensive. Results show that the proposed approach surpasses the state-of-art approaches by a significant margin.
Zero-shot classification is the task of classifying the test sample into a new class which was not seen during training. This is possible by establishing a relationship between the training and the testing classes using auxiliary information. In the second part of this thesis, a framework is designed that trains using the handcrafted attribute vectors and word vectors but doesn’t require the expensive attribute vectors during test time. More specifically, an intermediate space is learnt between the word vector space and the image feature space using the hand-crafted attribute vectors. Preliminary results on two zero-shot classification datasets show that this is a promising direction to explore.
ContributorsGattupalli, Jaya Vijetha (Author) / Li, Baoxin (Thesis advisor) / Yang, Yezhou (Committee member) / Venkateswara, Hemanth (Committee member) / Arizona State University (Publisher)
Created2019
Description
This research investigates the fine scale structure in Earth's mantle, especially for the lowermost mantle, where strong heterogeneity exists. Recent seismic tomography models have resolved large-scale features in the lower mantle, such as the large low shear velocity provinces (LLSVPs). However, differences are present between different models, especially at shorter length scales. Fine scale structures both within and outside LLSVPs are still poorly constrained. The drastic growth of global seismic networks presents densely sampled seismic data in unprecedented quality and quantity. In this work, the Empirical Wavelet construction method has been developed to document seismic travel time and waveform information for a global shear wave seismic dataset. A dataset of 250K high-quality seismic records with comprehensive measurements is documented and made publicly available. To more accurately classify high quality seismic signal from the noise, 1.4 million manually labeled seismic records have been used to train a supervised classification model. The constructed model performed better than the empirical model deployed in the Empirical Wavelet method, with 87% in precision and 83% in recall. To utilize lower amplitude phases such as higher multiples of S and ScS waves, we have developed a geographic bin stacking method to improve signal-to-noise ratio. It is then applied to Sn waves up to n=6 and ScSn wave up to n=5 for both minor and major arc phases. The virtual stations constructed provide unique path sampling and coverage, vastly improving sampling in the Southern Hemisphere. With the high-quality dataset we have gathered, ray-based layer stripping iterative forward tomography is implemented to update a starting tomography model by mapping the travel time residuals along the ray from the surface down to the core mantle boundary. Final updated models with different starting tomography models show consistent updates, suggesting a convergent solution. The final updated models show higher resolution results than the starting tomography models, especially on intermediate-scale structures. The combined analyses and results in this work provide new tools and new datasets to image the fine-scale heterogeneous structures in the lower mantle, which advances our understanding of the dynamics and evolution of the Earth's mantle.
ContributorsLai, Hongyu (Author) / Garnero, Edward J (Thesis advisor) / Till, Christy B. (Committee member) / Shim, Sang-Heon (Committee member) / Li, Mingming (Committee member) / Tyburczy, James (Committee member) / Arizona State University (Publisher)
Created2019