Matching Items (21)
Filtering by
- All Subjects: deep learning
- Creators: Turaga, Pavan
- Creators: School of Mathematical and Statistical Sciences
Description
Accurate pose initialization and pose estimation are crucial requirements in on-orbit space assembly and various other autonomous on-orbit tasks. However, pose initialization and pose estimation are much more difficult to do accurately and consistently in space. This is primarily due to not only the variable lighting conditions present in space, but also the power requirements mandated by space-flyable hardware. This thesis investigates leveraging a deep learning approach for monocular one-shot pose initialization and pose estimation. A convolutional neural network was used to estimate the 6D pose of an assembly truss object. This network was trained by utilizing synthetic imagery generated from a simulation testbed. Furthermore, techniques to quantify model uncertainty of the deep learning model were investigated and applied in the task of in-space pose estimation and pose initialization. The feasibility of this approach on low-power computational platforms was also tested. The results demonstrate that accurate pose initialization and pose estimation can be conducted using a convolutional neural network. In addition, the results show that the model uncertainty can be obtained from the network. Lastly, the use of deep learning for pose initialization and pose estimation in addition with uncertainty quantification was demonstrated to be feasible on low-power compute platforms.
ContributorsKailas, Siva Maneparambil (Author) / Ben Amor, Heni (Thesis director) / Detry, Renaud (Committee member) / Economics Program in CLAS (Contributor) / School of Mathematical and Statistical Sciences (Contributor) / Computer Science and Engineering Program (Contributor) / Barrett, The Honors College (Contributor)
Created2020-05
Description
In shotgun proteomics, liquid chromatography coupled to tandem mass spectrometry
(LC-MS/MS) is used to identify and quantify peptides and proteins. LC-MS/MS produces mass spectra, which must be searched by one or more engines, which employ
algorithms to match spectra to theoretical spectra derived from a reference database.
These engines identify and characterize proteins and their component peptides. By
training a convolutional neural network on a dataset of over 6 million MS/MS spectra
derived from human proteins, we aim to create a tool that can quickly and effectively
identify spectra as peptides prior to database searching. This can significantly reduce search space and thus run time for database searches, thereby accelerating LCMS/MS-based proteomics data acquisition. Additionally, by training neural networks
on labels derived from the search results of three different database search engines, we
aim to examine and compare which features are best identified by individual search
engines, a neural network, or a combination of these.
(LC-MS/MS) is used to identify and quantify peptides and proteins. LC-MS/MS produces mass spectra, which must be searched by one or more engines, which employ
algorithms to match spectra to theoretical spectra derived from a reference database.
These engines identify and characterize proteins and their component peptides. By
training a convolutional neural network on a dataset of over 6 million MS/MS spectra
derived from human proteins, we aim to create a tool that can quickly and effectively
identify spectra as peptides prior to database searching. This can significantly reduce search space and thus run time for database searches, thereby accelerating LCMS/MS-based proteomics data acquisition. Additionally, by training neural networks
on labels derived from the search results of three different database search engines, we
aim to examine and compare which features are best identified by individual search
engines, a neural network, or a combination of these.
ContributorsWhyte, Cameron Stafford (Author) / Suren, Jayasuriya (Thesis director) / Gil, Speyer (Committee member) / Patrick, Pirrotte (Committee member) / School of Mathematical and Statistical Sciences (Contributor, Contributor) / Barrett, The Honors College (Contributor)
Created2020-05
Description
Convolutional neural networks boast a myriad of applications in artificial intelligence, but one of the most common uses for such networks is image extraction. The ability of convolutional layers to extract and combine data features for the purpose of image analysis can be leveraged for pose estimation on an object - detecting the presence and attitude of corners and edges allows a convolutional neural network to identify how an object is positioned. This task can assist in working to grasp an object correctly in robotics applications, or to track an object more accurately in 3D space. However, the effectiveness of pose estimation may change based on properties of the object; the pose of a complex object, complexity being determined by internal occlusions, similar faces, etcetera, can be difficult to resolve.
This thesis is part of a collaboration between ASU’s Interactive Robotics Laboratory and NASA’s Jet Propulsion Laboratory. In this thesis, the training pipeline from Sharma’s paper “Pose Estimation for Non-Cooperative Spacecraft Rendezvous Using Convolutional Neural Networks” was modified to perform pose estimation on a complex object - specifically, a segment of a hollow truss. After initial attempts to replicate the architecture used in the paper and train solely on synthetic images, a combination of synthetic dataset generation and transfer learning on an ImageNet-pretrained AlexNet model was implemented to mitigate the difficulty of gathering large amounts of real-world data. Experimentation with pose estimation accuracy and hyperparameters of the model resulted in gradual test accuracy improvement, and future work is suggested to improve pose estimation for complex objects with some form of rotational symmetry.
This thesis is part of a collaboration between ASU’s Interactive Robotics Laboratory and NASA’s Jet Propulsion Laboratory. In this thesis, the training pipeline from Sharma’s paper “Pose Estimation for Non-Cooperative Spacecraft Rendezvous Using Convolutional Neural Networks” was modified to perform pose estimation on a complex object - specifically, a segment of a hollow truss. After initial attempts to replicate the architecture used in the paper and train solely on synthetic images, a combination of synthetic dataset generation and transfer learning on an ImageNet-pretrained AlexNet model was implemented to mitigate the difficulty of gathering large amounts of real-world data. Experimentation with pose estimation accuracy and hyperparameters of the model resulted in gradual test accuracy improvement, and future work is suggested to improve pose estimation for complex objects with some form of rotational symmetry.
ContributorsDsouza, Susanna Roshini (Author) / Ben Amor, Hani (Thesis director) / Maneparambil, Kailasnath (Committee member) / Computer Science and Engineering Program (Contributor, Contributor) / School of Mathematical and Statistical Sciences (Contributor) / Barrett, The Honors College (Contributor)
Created2019-05
Description
YouTube video bots have been constantly generating bot videos and posting them on the YouTube platform. While these bot-generated videos negatively influence the YouTube audience, they cost YouTube extra resources to host. The goal for this project is to build a classifier that identifies bot-generated channels based on a deep learning-based framework. We designed the framework to take text, audio, and video features into account. For the purpose of this thesis project, we will be focusing on text classification work.
ContributorsSai, Lun (Author) / Benjamin, Victor (Thesis director) / Lin, Elva S.Y. (Committee member) / Department of Information Systems (Contributor, Contributor) / School of Accountancy (Contributor) / School of Mathematical and Statistical Sciences (Contributor) / Barrett, The Honors College (Contributor)
Created2019-05
Description
This thesis encompasses a comprehensive research effort dedicated to overcoming the critical bottlenecks that hinder the current generation of neural networks, thereby significantly advancing their reliability and performance. Deep neural networks, with their millions of parameters, suffer from over-parameterization and lack of constraints, leading to limited generalization capabilities. In other words, the complex architecture and millions of parameters present challenges in finding the right balance between capturing useful patterns and avoiding noise in the data. To address these issues, this thesis explores novel solutions based on knowledge distillation, enabling the learning of robust representations. Leveraging the capabilities of large-scale networks, effective learning strategies are developed. Moreover, the limitations of dependency on external networks in the distillation process, which often require large-scale models, are effectively overcome by proposing a self-distillation strategy. The proposed approach empowers the model to generate high-level knowledge within a single network, pushing the boundaries of knowledge distillation. The effectiveness of the proposed method is not only demonstrated across diverse applications, including image classification, object detection, and semantic segmentation but also explored in practical considerations such as handling data scarcity and assessing the transferability of the model to other learning tasks. Another major obstacle hindering the development of reliable and robust models lies in their black-box nature, impeding clear insights into the contributions toward the final predictions and yielding uninterpretable feature representations. To address this challenge, this thesis introduces techniques that incorporate simple yet powerful deep constraints rooted in Riemannian geometry. These constraints confer geometric qualities upon the latent representation, thereby fostering a more interpretable and insightful representation. In addition to its primary focus on general tasks like image classification and activity recognition, this strategy offers significant benefits in real-world applications where data scarcity is prevalent. Moreover, its robustness in feature removal showcases its potential for edge applications. By successfully tackling these challenges, this research contributes to advancing the field of machine learning and provides a foundation for building more reliable and robust systems across various application domains.
ContributorsChoi, Hongjun (Author) / Turaga, Pavan (Thesis advisor) / Jayasuriya, Suren (Committee member) / Li, Wenwen (Committee member) / Fazli, Pooyan (Committee member) / Arizona State University (Publisher)
Created2023
Description
The study tested the parameterized neural ordinary differential equation (PNODE) framework with a physical system exhibiting only advective phenomenon. Existing deep learning methods have difficulty learning multiple dynamic, continuous time processes. PNODE encodes the input data and initial parameter into a set of reduced states within the latent space. Then the reduced states are fitted to a system of ordinary differential equations. The outputs from the model are then decoded back to the data space for a desired input parameter and time. The application of the PNODE formalism to different types of physical systems is important to test the methods robustness. The linear advection data was generated through a high-fidelity numerical tool for multiple velocity parameters. The PNODE code was modified for the advection dataset, whose temporal domain and spatial discretization varied from the original study configuration. The L2 norm between the reconstruction and surrogate model and the reconstruction plots were used to analyze the PNODE model performance. The model reconstructions presented mixed results. For a temporal domain of 20-time units, where multiple advection cycles were completed for each advection speed, the reconstructions did not agree with the surrogate model. For a reduced temporal domain of 5-time units, the reconstructions and surrogate models were in close agreement. Near the end of the temporal domain, deviations occurred likely resulting from the accumulation of numerical errors. Note, over the 5-time units, smaller advection speed parameters were unable to complete a cycle. The behavior for the 20-time units highlighted potential issues with imbalanced datasets and repeated features. The 5-time unit model illustrates PNODEs adaptability to this class of problems when the dataset is better posed.
ContributorsReithal, Richard Robert (Author) / Kim, Jeonglae (Thesis director) / Lee, Kookjin (Committee member) / Barrett, The Honors College (Contributor) / Mechanical and Aerospace Engineering Program (Contributor) / School of Mathematical and Statistical Sciences (Contributor)
Created2022-12
Description
Breast cancer is one of the most common types of cancer worldwide. Early detection and diagnosis are crucial for improving the chances of successful treatment and survival. In this thesis, many different machine learning algorithms were evaluated and compared to predict breast cancer malignancy from diagnostic features extracted from digitized images of breast tissue samples, called fine-needle aspirates. Breast cancer diagnosis typically involves a combination of mammography, ultrasound, and biopsy. However, machine learning algorithms can assist in the detection and diagnosis of breast cancer by analyzing large amounts of data and identifying patterns that may not be discernible to the human eye. By using these algorithms, healthcare professionals can potentially detect breast cancer at an earlier stage, leading to more effective treatment and better patient outcomes. The results showed that the gradient boosting classifier performed the best, achieving an accuracy of 96% on the test set. This indicates that this algorithm can be a useful tool for healthcare professionals in the early detection and diagnosis of breast cancer, potentially leading to improved patient outcomes.
ContributorsMallya, Aatmik (Author) / De Luca, Gennaro (Thesis director) / Chen, Yinong (Committee member) / Barrett, The Honors College (Contributor) / School of Mathematical and Statistical Sciences (Contributor) / Computer Science and Engineering Program (Contributor)
Created2023-05
Description
Hydrologic modeling in snowfed karst watersheds is important for many communities relying on their water for municipal and agricultural use, but the complexities of karst hydrology have made this task historically difficult. Here, two Long Short-Term Memory (LSTM) models are compared to investigate this problem from a deep-learning perspective within the context of the Logan River Canyon watershed, which supplies water to Logan City, UT. One is spatially lumped and the other spatially distributed, the latter with a potential to reveal underlying spatial watershed dynamics. Both use snowmelt and rainfall to predict daily streamflow downstream. I find distributed LSTMs consistently outperform lumped LSTMs in this task. Additionally, I find that a spatial sensitivity analysis of distributed LSTMs is unpromising in revealing spatial watershed dynamics but warrants further investigation.
ContributorsShaver, Ryan (Author) / Xu, Tianfang (Thesis director) / Jones, Don (Committee member) / Barrett, The Honors College (Contributor) / School of Earth and Space Exploration (Contributor) / School of Mathematical and Statistical Sciences (Contributor)
Created2022-05
Description
Millimeter-wave (mmWave) and sub-terahertz (sub-THz) systems aim to utilize the large bandwidth available at these frequencies. This has the potential to enable several future applications that require high data rates, such as autonomous vehicles and digital twins. These systems, however, have several challenges that need to be addressed to realize their gains in practice. First, they need to deploy large antenna arrays and use narrow beams to guarantee sufficient receive power. Adjusting the narrow beams of the large antenna arrays incurs massive beam training overhead. Second, the sensitivity to blockages is a key challenge for mmWave and THz networks. Since these networks mainly rely on line-of-sight (LOS) links, sudden link blockages highly threaten the reliability of the networks. Further, when the LOS link is blocked, the network typically needs to hand off the user to another LOS basestation, which may incur critical time latency, especially if a search over a large codebook of narrow beams is needed. A promising way to tackle both these challenges lies in leveraging additional side information such as visual, LiDAR, radar, and position data. These sensors provide rich information about the wireless environment, which can be utilized for fast beam and blockage prediction. This dissertation presents a machine-learning framework for sensing-aided beam and blockage prediction. In particular, for beam prediction, this work proposes to utilize visual and positional data to predict the optimal beam indices. For the first time, this work investigates the sensing-aided beam prediction task in a real-world vehicle-to-infrastructure and drone communication scenario. Similarly, for blockage prediction, this dissertation proposes a multi-modal wireless communication solution that utilizes bimodal machine learning to perform proactive blockage prediction and user hand-off. Evaluations on both real-world and synthetic datasets illustrate the promising performance of the proposed solutions and highlight their potential for next-generation communication and sensing systems.
ContributorsCharan, Gouranga (Author) / Alkhateeb, Ahmed (Thesis advisor) / Chakrabarti, Chaitali (Committee member) / Turaga, Pavan (Committee member) / Michelusi, Nicolò (Committee member) / Arizona State University (Publisher)
Created2024
Description
Mixture of experts is a machine learning ensemble approach that consists of individual models that are trained to be ``experts'' on subsets of the data, and a gating network that provides weights to output a combination of the expert predictions. Mixture of experts models do not currently see wide use due to difficulty in training diverse experts and high computational requirements. This work presents modifications of the mixture of experts formulation that use domain knowledge to improve training, and incorporate parameter sharing among experts to reduce computational requirements.
First, this work presents an application of mixture of experts models for quality robust visual recognition. First it is shown that human subjects outperform deep neural networks on classification of distorted images, and then propose a model, MixQualNet, that is more robust to distortions. The proposed model consists of ``experts'' that are trained on a particular type of image distortion. The final output of the model is a weighted sum of the expert models, where the weights are determined by a separate gating network. The proposed model also incorporates weight sharing to reduce the number of parameters, as well as increase performance.
Second, an application of mixture of experts to predict visual saliency is presented. A computational saliency model attempts to predict where humans will look in an image. In the proposed model, each expert network is trained to predict saliency for a set of closely related images. The final saliency map is computed as a weighted mixture of the expert networks' outputs, with weights determined by a separate gating network. The proposed model achieves better performance than several other visual saliency models and a baseline non-mixture model.
Finally, this work introduces a saliency model that is a weighted mixture of models trained for different levels of saliency. Levels of saliency include high saliency, which corresponds to regions where almost all subjects look, and low saliency, which corresponds to regions where some, but not all subjects look. The weighted mixture shows improved performance compared with baseline models because of the diversity of the individual model predictions.
First, this work presents an application of mixture of experts models for quality robust visual recognition. First it is shown that human subjects outperform deep neural networks on classification of distorted images, and then propose a model, MixQualNet, that is more robust to distortions. The proposed model consists of ``experts'' that are trained on a particular type of image distortion. The final output of the model is a weighted sum of the expert models, where the weights are determined by a separate gating network. The proposed model also incorporates weight sharing to reduce the number of parameters, as well as increase performance.
Second, an application of mixture of experts to predict visual saliency is presented. A computational saliency model attempts to predict where humans will look in an image. In the proposed model, each expert network is trained to predict saliency for a set of closely related images. The final saliency map is computed as a weighted mixture of the expert networks' outputs, with weights determined by a separate gating network. The proposed model achieves better performance than several other visual saliency models and a baseline non-mixture model.
Finally, this work introduces a saliency model that is a weighted mixture of models trained for different levels of saliency. Levels of saliency include high saliency, which corresponds to regions where almost all subjects look, and low saliency, which corresponds to regions where some, but not all subjects look. The weighted mixture shows improved performance compared with baseline models because of the diversity of the individual model predictions.
ContributorsDodge, Samuel Fuller (Author) / Karam, Lina (Thesis advisor) / Jayasuriya, Suren (Committee member) / Li, Baoxin (Committee member) / Turaga, Pavan (Committee member) / Arizona State University (Publisher)
Created2018