Matching Items (92)
Filtering by
- Genre: Masters Thesis
- Creators: Li, Baoxin
- Creators: Cooke, Nancy J.
Description
Mindfulness meditation practices have become an intervention of focus in the literature, but little attention has been drawn to the effectiveness of this practice as a single execution in an online format. Several approaches were employed to capture the effects of a mindful breathing exercise and yoga experience on working memory capacity. Through several analyses, they found that though there was no significant difference between working memory capacity scores before and after this breathing exercise, and mindfulness and yoga experience had no influence on working memory performance. Although these findings were not statistically significant, there are several trends to note and implications for this research within the body of literature.
ContributorsSpecht, Rachel A (Author) / Becker, Vaughn (Thesis advisor) / Black, Candace (Committee member) / Cooke, Nancy J. (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
The current study aims to explore factors affecting trust in human-drone collaboration. A current gap exists in research surrounding civilian drone use and the role of trust in human-drone interaction and collaboration. Specifically, existing research lacks an explanation of the relationship between drone pilot experience, trust, and trust-related behaviors as well as other factors. Using two dimensions of trust in human-automation team—purpose and performance—the effects of experience on drone design and trust is studied to explore factors that may contribute to such a model. An online survey was conducted to examine civilian drone operators’ experience, familiarity, expertise, and trust in commercially available drones. It was predicted that factors of prior experience (familiarity, self-reported expertise) would have a significant effect on trust in drones. The choice to use or exclude the drone propellers in a search-and-identify scenario, paired with the pilots’ experience with drones, would further confirm the relevance of the trust dimensions of purpose versus performance in the human-drone relationship. If the pilot has a positive sense of purpose and benevolence with the drone, the pilot trusts the drone has a positive intent towards them and the task. If the pilot has trust in the performance of the drone, they ascertain that the drone has the skill to do the task. The researcher found no significant differences between mean trust scores across levels of familiarity, but did find some interaction between self-report expertise, familiarity, and trust. Future research should further explore more concrete measures of situational participant factors such as self-confidence and expertise to understand their role in civilian pilots’ trust in their drone.
ContributorsNiichel, Madeline Kathleen (Author) / Chiou, Erin (Thesis advisor) / Cooke, Nancy J. (Committee member) / Craig, Scotty (Committee member) / Arizona State University (Publisher)
Created2019
Description
In recent years, Convolutional Neural Networks (CNNs) have been widely used in not only the computer vision community but also within the medical imaging community. Specifically, the use of pre-trained CNNs on large-scale datasets (e.g., ImageNet) via transfer learning for a variety of medical imaging applications, has become the de facto standard within both communities.
However, to fit the current paradigm, 3D imaging tasks have to be reformulated and solved in 2D, losing rich 3D contextual information. Moreover, pre-trained models on natural images never see any biomedical images and do not have knowledge about anatomical structures present in medical images. To overcome the above limitations, this thesis proposes an image out-painting self-supervised proxy task to develop pre-trained models directly from medical images without utilizing systematic annotations. The idea is to randomly mask an image and train the model to predict the missing region. It is demonstrated that by predicting missing anatomical structures when seeing only parts of the image, the model will learn generic representation yielding better performance on various medical imaging applications via transfer learning.
The extensive experiments demonstrate that the proposed proxy task outperforms training from scratch in six out of seven medical imaging applications covering 2D and 3D classification and segmentation. Moreover, image out-painting proxy task offers competitive performance to state-of-the-art models pre-trained on ImageNet and other self-supervised baselines such as in-painting. Owing to its outstanding performance, out-painting is utilized as one of the self-supervised proxy tasks to provide generic 3D pre-trained models for medical image analysis.
However, to fit the current paradigm, 3D imaging tasks have to be reformulated and solved in 2D, losing rich 3D contextual information. Moreover, pre-trained models on natural images never see any biomedical images and do not have knowledge about anatomical structures present in medical images. To overcome the above limitations, this thesis proposes an image out-painting self-supervised proxy task to develop pre-trained models directly from medical images without utilizing systematic annotations. The idea is to randomly mask an image and train the model to predict the missing region. It is demonstrated that by predicting missing anatomical structures when seeing only parts of the image, the model will learn generic representation yielding better performance on various medical imaging applications via transfer learning.
The extensive experiments demonstrate that the proposed proxy task outperforms training from scratch in six out of seven medical imaging applications covering 2D and 3D classification and segmentation. Moreover, image out-painting proxy task offers competitive performance to state-of-the-art models pre-trained on ImageNet and other self-supervised baselines such as in-painting. Owing to its outstanding performance, out-painting is utilized as one of the self-supervised proxy tasks to provide generic 3D pre-trained models for medical image analysis.
ContributorsSodha, Vatsal Arvindkumar (Author) / Liang, Jianming (Thesis advisor) / Devarakonda, Murthy (Committee member) / Li, Baoxin (Committee member) / Arizona State University (Publisher)
Created2020
Description
Convolutional Neural Network (CNN) has achieved state-of-the-art performance in numerous applications like computer vision, natural language processing, robotics etc. The advancement of High-Performance Computing systems equipped with dedicated hardware accelerators has also paved the way towards the success of compute intensive CNNs. Graphics Processing Units (GPUs), with massive processing capability, have been of general interest for the acceleration of CNNs. Recently, Field Programmable Gate Arrays (FPGAs) have been promising in CNN acceleration since they offer high performance while also being re-configurable to support the evolution of CNNs. This work focuses on a design methodology to accelerate CNNs on FPGA with low inference latency and high-throughput which are crucial for scenarios like self-driving cars, video surveillance etc. It also includes optimizations which reduce the resource utilization by a large margin with a small degradation in performance thus making the design suitable for low-end FPGA devices as well.
FPGA accelerators often suffer due to the limited main memory bandwidth. Also, highly parallel designs with large resource utilization often end up achieving low operating frequency due to poor routing. This work employs data fetch and buffer mechanisms, designed specifically for the memory access pattern of CNNs, that overlap computation with memory access. This work proposes a novel arrangement of the systolic processing element array to achieve high frequency and consume less resources than the existing works. Also, support has been extended to more complicated CNNs to do video processing. On Intel Arria 10 GX1150, the design operates at a frequency as high as 258MHz and performs single inference of VGG-16 and C3D in 23.5ms and 45.6ms respectively. For VGG-16 and C3D the design offers a throughput of 66.1 and 23.98 inferences/s respectively. This design can outperform other FPGA 2D CNN accelerators by up to 9.7 times and 3D CNN accelerators by up to 2.7 times.
FPGA accelerators often suffer due to the limited main memory bandwidth. Also, highly parallel designs with large resource utilization often end up achieving low operating frequency due to poor routing. This work employs data fetch and buffer mechanisms, designed specifically for the memory access pattern of CNNs, that overlap computation with memory access. This work proposes a novel arrangement of the systolic processing element array to achieve high frequency and consume less resources than the existing works. Also, support has been extended to more complicated CNNs to do video processing. On Intel Arria 10 GX1150, the design operates at a frequency as high as 258MHz and performs single inference of VGG-16 and C3D in 23.5ms and 45.6ms respectively. For VGG-16 and C3D the design offers a throughput of 66.1 and 23.98 inferences/s respectively. This design can outperform other FPGA 2D CNN accelerators by up to 9.7 times and 3D CNN accelerators by up to 2.7 times.
ContributorsRavi, Pravin Kumar (Author) / Zhao, Ming (Thesis advisor) / Li, Baoxin (Committee member) / Ren, Fengbo (Committee member) / Arizona State University (Publisher)
Created2020
Description
Communications between air traffic controllers and pilots are critical to national airspace traffic management. Measuring communications in real time made by pilots and air traffic controllers has the potential to predict human error. In this thesis a measure for Deviations from Closed Loop Communications is defined and tested to predict a human error event, Loss of Separation (LOS). Six retired air traffic controllers were recruited and tested in three conditions of varying workload in an Terminal Radar Approach Control Facility (TRACON) arrival radar simulation. Communication transcripts from simulated trials were transcribed and coding schemes for Closed Loop Communication Deviations (CLCD) were applied. Results of the study demonstrated a positive correlation between CLCD and LOS, indicating that CLCD could be a variable used to predict LOS. However, more research is required to determine if CLCD can be used to predict LOS independent of other predictor variables, and if CLCD can be used in a model that considers many different predictor variables to predict LOS.
ContributorsLieber, Christopher Shane (Author) / Cooke, Nancy J. (Thesis advisor) / Gutzwiller, Robert S (Committee member) / Niemczyk, Mary (Committee member) / Arizona State University (Publisher)
Created2020
Description
Analytics are being collected on a day to day basis on just about anything that you can think of. Sports is one of the recent fields that has started implementing the tool into their game. Analytics can be described as an abundance of statistical information that show situational tendencies of other teams and players. It is hypothesized that analytics provide anticipatory information that allows athletes to know what is coming; therefore, allowing them to perform better in real game scenarios. However, it is unclear how this information should be presented to athletes and whether athletes can actually retain the abundance of information given to them. Two different types of presentation methods (Numeric and Numeric plus Graph) and two different amounts of analytic information (High and Low) were compared for baseball players in an online based baseball specific retention survey: High Numeric (excess information shown in spreadsheet format), Low Numeric (key information shown in spreadsheet format), High Numeric plus Graph (excess information shown as a spreadsheet with hitting zone maps), and Low Numeric plus Graph (key information shown as a spreadsheet with hitting zone maps). Athletes produced different retention scores for the type of presentation method given across the whole study. Athletes presented analytic as Numeric plus Graph performed better than athletes in just Numeric condition. Additionally, playing experience had a significant effect on an athlete’s ability to retain analytic information. Athletes with 10 plus years of baseball experience performed better in every condition other than High Numeric plus Graph compared to athletes with less than 10 years of experience. Amount and experience also had an interaction effect that produced statistical significance; those with less experience performed better in conditions with less baseball information given whereas those with more experience were able to handle more baseball information at once. Providing analytic information gives athletes, especially baseball batters, a significant advantage over their opponent; however, ability to retain analytic information depends on how the information is presented and to whom the information is being presented.
ContributorsGin, Andrew B (Author) / Gray, Robert (Thesis advisor) / Cooke, Nancy J. (Committee member) / Craig, Scotty (Committee member) / Arizona State University (Publisher)
Created2020
Description
The development of the internet provided new means for people to communicate effectively and share their ideas. There has been a decline in the consumption of newspapers and traditional broadcasting media toward online social mediums in recent years. Social media has been introduced as a new way of increasing democratic discussions on political and social matters. Among social media, Twitter is widely used by politicians, government officials, communities, and parties to make announcements and reach their voice to their followers. This greatly increases the acceptance domain of the medium.
The usage of social media during social and political campaigns has been the subject of a lot of social science studies including the Occupy Wall Street movement, The Arab Spring, the United States (US) election, more recently The Brexit campaign. The wide
spread usage of social media in this space and the active participation of people in the discussions on social media made this communication channel a suitable place for spreading propaganda to alter public opinion.
An interesting feature of twitter is the feasibility of which bots can be programmed to operate on this platform. Social media bots are automated agents engineered to emulate the activity of a human being by tweeting some specific content, replying to users, magnifying certain topics by retweeting them. Network on these bots is called botnets and describing the collaboration of connected computers with programs that communicates across multiple devices to perform some task.
In this thesis, I will study how bots can influence the opinion, finding which parameters are playing a role in shrinking or coalescing the communities, and finally logically proving the effectiveness of each of the hypotheses.
The usage of social media during social and political campaigns has been the subject of a lot of social science studies including the Occupy Wall Street movement, The Arab Spring, the United States (US) election, more recently The Brexit campaign. The wide
spread usage of social media in this space and the active participation of people in the discussions on social media made this communication channel a suitable place for spreading propaganda to alter public opinion.
An interesting feature of twitter is the feasibility of which bots can be programmed to operate on this platform. Social media bots are automated agents engineered to emulate the activity of a human being by tweeting some specific content, replying to users, magnifying certain topics by retweeting them. Network on these bots is called botnets and describing the collaboration of connected computers with programs that communicates across multiple devices to perform some task.
In this thesis, I will study how bots can influence the opinion, finding which parameters are playing a role in shrinking or coalescing the communities, and finally logically proving the effectiveness of each of the hypotheses.
ContributorsAhmadi, Mohsen (Author) / Davulcu, Hasan (Thesis advisor) / Sen, Arunabha (Committee member) / Li, Baoxin (Committee member) / Arizona State University (Publisher)
Created2020
Description
The choices of an operator under heavy cognitive load are potentially critical to overall safety and performance. Such conditions are common when technological failures arise, and the operator is forced into multi-task situations. Task switching choice was examined in an effort to both validate previous work concerning a model of task overload management and address unresolved matters related to visual sampling. Using the Multi-Attribute Task Battery and eye tracking, the experiment studied any influence of task priority and difficulty. Continuous visual attention measurements captured attentional switches that do not manifest into behaviors but may provide insight into task switching choice. Difficulty was found to have an influence on task switching behavior; however, priority was not. Instead, priority may affect time spent on a task rather than strictly choice. Eye measures revealed some moderate connections between time spent dwelling on a task and subjective interest. The implication of this, as well as eye tracking used to validate a model of task overload management as a whole, is discussed.
ContributorsZabala, Garrett (Author) / Gutzwiller, Robert S (Thesis advisor) / Cooke, Nancy J. (Committee member) / Gray, Rob (Committee member) / Arizona State University (Publisher)
Created2020
Description
Despite the prevalence of teams in complex sociotechnical systems, current approaches to understanding workload tend to focus on the individual operator. However, research suggests that team workload has emergent properties and is not necessarily equivalent to the aggregate of individual workload. Assessment of communications provides a means of examining aspects of team workload in highly interdependent teams. This thesis set out to explore how communications are associated with team workload and performance under high task demand in all-human and human–autonomy teams in a command and control task. A social network analysis approach was used to analyze the communications of 30 different teams, each with three members operating in a command and control task environment of over a series of five missions. Teams were assigned to conditions differentiated by their composition with either a naïve participant, a trained confederate, or a synthetic agent in the pilot role. Social network analysis measures of centralization and intensity were used to assess differences in communications between team types and under different levels of demand, and relationships between communication measures, performance, and workload distributions were also examined. Results indicated that indegree centralization was greater in the all-human control teams than in the other team types, but degree centrality standard deviation and intensity were greatest in teams with a highly trained experimenter pilot. In all three team types, the intensity of communications and degree centrality standard deviation appeared to decrease during the high demand mission, but indegree and outdegree centralization did not. Higher communication intensity was associated with more efficient target processing and more successful target photos per mission, but a clear relationship between measures of performance and decentralization of communications was not found.
ContributorsJohnson, Craig Jonathon (Author) / Cooke, Nancy J. (Thesis advisor) / Gray, Robert (Committee member) / Gutzwiller, Robert S (Committee member) / Arizona State University (Publisher)
Created2020