Matching Items (185)
Filtering by
- Genre: Academic theses
- Creators: Turaga, Pavan
Description
This thesis presents robust and novel solutions using knowledge distillation with geometric approaches and multimodal data that can address the current challenges in deep learning, providing a comprehensive understanding of the learning process involved in knowledge distillation. Deep learning has attained significant success in various applications, such as health and wellness promotion, smart homes, and intelligent surveillance. In general, stacking more layers or increasing the number of trainable parameters causes deep networks to exhibit improved performance. However, this causes the model to become large, resulting in an additional need for computing and power resources for training, storage, and deployment. These are the core challenges in incorporating such models into small devices with limited power and computational resources. In this thesis, robust solutions aimed at addressing the aforementioned challenges are presented. These proposed methodologies and algorithmic contributions enhance the performance and efficiency of deep learning models. The thesis encompasses a comprehensive exploration of knowledge distillation, an approach that holds promise for creating compact models from high-capacity ones, while preserving their performance. This exploration covers diverse datasets, including both time series and image data, shedding light on the pivotal role of augmentation methods in knowledge distillation. The effects of these methods are rigorously examined through empirical experiments. Furthermore, the study within this thesis delves into the efficient utilization of features derived from two different teacher models, each trained on dissimilar data representations, including time-series and image data. Through these investigations, I present novel approaches to knowledge distillation, leveraging geometric techniques for the analysis of multimodal data. These solutions not only address real-world challenges but also offer valuable insights and recommendations for modeling in new applications.
ContributorsJeon, Eunsom (Author) / Turaga, Pavan (Thesis advisor) / Li, Baoxin (Committee member) / Lee, Hyunglae (Committee member) / Jayasuriya, Suren (Committee member) / Arizona State University (Publisher)
Created2023
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
Although previous studies have elucidated the role of position feedback in the regulation of movement, the specific contribution of Golgi tendon organs (GTO) in force feedback, especially in stabilizing voluntary limb movements, has remained theoretical due to limitations in experimental techniques. This study aims to establish force feedback regulation mediated by GTO afferent signals in two phases. The first phase of this study consisted of simulations using a neuromusculoskeletal model of the monoarticular elbow flexor (MEF) muscle group, assess the impact of force feedback in maintaining steady state interaction forces against variable environmental stiffness. Three models were trained to accurately reach an interaction force of 40N, 50N and 60N respectively, using a fixed stiffness level. Next, the environment stiffness was switched between untrained levels for open loop (OL) and closed loop (CL) variants of the same model. Results showed that compared to OL, CL showed decreased force deviations by 10.43%, 12.11% and 13.02% for each of the models. Most importantly, it is also observed that in the absence of force feedback, environment stiffness is found to have an effect on the interaction force. In the second phase, human subjects were engaged in experiments utilizing an instrumented elbow exoskeleton that applied loads to the MEF muscle group, closely mimicking the simulation conditions. The experiments consisted of reference, blind and catch trial types, and 3 stiffness levels. Subjects were first trained to reach for a predetermined target force. During catch trials, stiffness levels were randomized between reaches. Responses obtained from these experiments showed that subjects were able to regulate forces with no significant effects of trial type or stiffness level. Since experimental results align closely with that of closed loop model simulations, the presence of force feedback mechanisms mediated by GTO within the human neuromuscular system is established. This study not only unveils the critical involvement of GTO in force feedback but also emphasizes the importance of understanding these mechanisms for developing advanced neuroprosthetics and rehabilitation strategies, shedding light on the intricate interplay between sensory inputs and motor responses in human proprioception.
ContributorsAbishek, Kevin (Author) / Lee, Hyunglae (Thesis advisor) / Buneo, Christopher (Committee member) / Santello, Marco (Committee member) / Arizona State University (Publisher)
Created2023
Description
Generative models are deep neural network-based models trained to learn the underlying distribution of a dataset. Once trained, these models can be used to sample novel data points from this distribution. Their impressive capabilities have been manifested in various generative tasks, encompassing areas like image-to-image translation, style transfer, image editing, and more. One notable application of generative models is data augmentation, aimed at expanding and diversifying the training dataset to augment the performance of deep learning models for a downstream task. Generative models can be used to create new samples similar to the original data but with different variations and properties that are difficult to capture with traditional data augmentation techniques. However, the quality, diversity, and controllability of the shape and structure of the generated samples from these models are often directly proportional to the size and diversity of the training dataset. A more extensive and diverse training dataset allows the generative model to capture overall structures present in the data and generate more diverse and realistic-looking samples.
In this dissertation, I present innovative methods designed to enhance the robustness and controllability of generative models, drawing upon physics-based, probabilistic, and geometric techniques. These methods help improve the generalization and controllability of the generative model without necessarily relying on large training datasets. I enhance the robustness of generative models by integrating classical geometric moments for shape awareness and minimizing trainable parameters. Additionally, I employ non-parametric priors for the generative model's latent space through basic probability and optimization methods to improve the fidelity of interpolated images. I adopt a hybrid approach to address domain-specific challenges with limited data and controllability, combining physics-based rendering with generative models for more realistic results.
These approaches are particularly relevant in industrial settings, where the training datasets are small and class imbalance is common. Through extensive experiments on various datasets, I demonstrate the effectiveness of the proposed methods over conventional approaches.
ContributorsSingh, Rajhans (Author) / Turaga, Pavan (Thesis advisor) / Jayasuriya, Suren (Committee member) / Berisha, Visar (Committee member) / Fazli, Pooyan (Committee member) / Arizona State University (Publisher)
Created2023
Description
A current thrust in neurorehabilitation research involves exogenous neuromodulation of peripheral nerves to enhance neuroplasticity and maximize recovery of function. This dissertation presents the results of four experiments aimed at assessing the effects of trigeminal nerve stimulation (TNS) and occipital nerve stimulation (ONS) on motor learning, which was behaviorally characterized using an upper extremity visuomotor adaptation paradigm. In Aim 1a, the effects of offline TNS using clinically tested frequencies (120 and 60 Hz) were characterized. Sixty-three participants (22.75±4.6 y/o), performed a visuomotor rotation task and received TNS before encountering rotation of hand visual feedback. In Aim 1b, TNS at 3 kHz, which has been shown to be more tolerable at higher current intensities, was evaluated in 42 additional subjects (23.4±4.6 y/o). Results indicated that 3 kHz stimulation accelerated learning while 60 Hz stimulation slowed learning, suggesting a frequency-dependent effect on learning. In Aim 2, the effect of online TNS using 120 and 60 Hz were characterized to determine if this protocol would deliver better outcomes. Sixty-three participants (23.2±3.9 y/o) received either TNS or sham concurrently with perturbed visual feedback. Results showed no significant differences among groups. However, a cross-study comparison of results obtained with 60 Hz offline TNS showed a statistically significant improvement in learning rates with online stimulation relative to offline, suggesting a timing-dependent effect on learning. In Aim 3, TNS and ONS were compared using the best protocol from previous aims (offline 3 kHz). Additionally, concurrent stimulation of both nerves was explored to look for potential synergistic effects. Eighty-four participants (22.9±3.2 y/o) were assigned to one of four groups: TNS, ONS, TNS+ONS, and sham. Visual inspection of learning curves revealed that the ONS group demonstrated the fastest learning among groups. However, statistical analyses did not confirm this observation. In addition, the TNS+ONS group appeared to learn faster than the sham and TNS groups but slower than the ONS only group, suggesting no synergistic effects using this protocol, as initially hypothesized.
The results provide new information on the potential use of TNS and ONS in neurorehabilitation and performance enhancement in the motor domain.
ContributorsArias, Diego (Author) / Buneo, Christopher (Thesis advisor) / Schaefer, Sydney (Committee member) / Helms-Tillery, Stephen (Committee member) / Santello, Marco (Committee member) / Kleim, Jeffrey (Committee member) / Arizona State University (Publisher)
Created2023
Description
Multisensory integration is the process by which information from different sensory modalities is integrated by the nervous system. This process is important not only from a basic science perspective but also for translational reasons, e.g., for the development of closed-loop neural prosthetic systems. A mixed virtual reality platform was developed to study the neural mechanisms of multisensory integration for the upper limb during motor planning. The platform allows for selection of different arms and manipulation of the locations of physical and virtual target cues in the environment. The system was tested with two non-human primates (NHP) trained to reach to multiple virtual targets. Arm kinematic data as well as neural spiking data from primary motor (M1) and dorsal premotor cortex (PMd) were collected. The task involved manipulating visual information about initial arm position by rendering the virtual avatar arm in either its actual position (veridical (V) condition) or in a different shifted (e.g., small vs large shifts) position (perturbed (P) condition) prior to movement. Tactile feedback was modulated in blocks by placing or removing the physical start cue on the table (tactile (T), and no-tactile (NT) conditions, respectively). Behaviorally, errors in initial movement direction were larger when the physical start cue was absent. Slightly larger directional errors were found in the P condition compared to the V condition for some movement directions. Both effects were consistent with the idea that erroneous or reduced information about initial hand location led to movement direction-dependent reach planning errors. Neural correlates of these behavioral effects were probed using population decoding techniques. For small shifts in the visual position of the arm, no differences in decoding accuracy between the T and NT conditions were observed in either M1 or PMd. However, for larger visual shifts, decoding accuracy decreased in the NT condition, but only in PMd. Thus, activity in PMd, but not M1, may reflect the uncertainty in reach planning that results when sensory cues regarding initial hand position are erroneous or absent.
ContributorsPhataraphruk, Preyaporn Kris (Author) / Buneo, Christopher A (Thesis advisor) / Zhou, Yi (Committee member) / Helms Tillery, Steve (Committee member) / Greger, Bradley (Committee member) / Santello, Marco (Committee member) / Arizona State University (Publisher)
Created2023
Description
In the standard pipeline for machine learning model development, several design decisions are made largely based on trial and error. Take the classification problem as an example. The starting point for classifier design is a dataset with samples from the classes of interest. From this, the algorithm developer must decide which features to extract, which hypothesis class to condition on, which hyperparameters to select, and how to train the model. The design process is iterative with the developer trying different classifiers, feature sets, and hyper-parameters and using cross-validation to pick the model with the lowest error. As there are no guidelines for when to stop searching, developers can continue "optimizing" the model to the point where they begin to "fit to the dataset". These problems are amplified in the active learning setting, where the initial dataset may be unlabeled and label acquisition is costly. The aim in this dissertation is to develop algorithms that provide ML developers with additional information about the complexity of the underlying problem to guide downstream model development. I introduce the concept of "meta-features" - features extracted from a dataset that characterize the complexity of the underlying data generating process. In the context of classification, the complexity of the problem can be characterized by understanding two complementary meta-features: (a) the amount of overlap between classes, and (b) the geometry/topology of the decision boundary. Across three complementary works, I present a series of estimators for the meta-features that characterize overlap and geometry/topology of the decision boundary, and demonstrate how they can be used in algorithm development.
ContributorsLi, Weizhi (Author) / Berisha, Visar (Thesis advisor) / Dasarathy, Gautam (Thesis advisor) / Natesan Ramamurthy, Karthikeyan (Committee member) / Turaga, Pavan (Committee member) / Arizona State University (Publisher)
Created2022
Description
With an aging population, the number of later in life health related incidents like stroke stand to become more prevalent. Unfortunately, the majority those who are most at risk for debilitating heath episodes are either uninsured or under insured when it comes to long term physical/occupational therapy. As insurance companies lower coverage and/or raise prices of plans with sufficient coverage, it can be expected that the proportion of uninsured/under insured to fully insured people will rise. To address this, lower cost alternative methods of treatment must be developed so people can obtain the treated required for a sufficient recovery. The presented robotic glove employs low cost fabric soft pneumatic actuators which use a closed loop feedback controller based on readings from embedded soft sensors. This provides the device with proprioceptive abilities for the dynamic control of each independent actuator. Force and fatigue tests were performed to determine the viability of the actuator design. A Box and Block test along with a motion capture study was completed to study the performance of the device. This paper presents the design and classification of a soft robotic glove with a feedback controller as a at-home stroke rehabilitation device.
ContributorsAxman, Reed C (Author) / Zhang, Wenlong (Thesis advisor) / Santello, Marco (Committee member) / McDaniel, Troy (Committee member) / Arizona State University (Publisher)
Created2022
Description
Severe forms of mental illness, such as schizophrenia and bipolar disorder, are debilitating conditions that negatively impact an individual's quality of life. Additionally, they are often difficult and expensive to diagnose and manage, placing a large burden on society. Mental illness is typically diagnosed by the use of clinical interviews and a set of neuropsychiatric batteries; a key component of nearly all of these evaluations is some spoken language task. Clinicians have long used speech and language production as a proxy for neurological health, but most of these assessments are subjective in nature. Meanwhile, technological advancements in speech and natural language processing have grown exponentially over the past decade, increasing the capacity of computer models to assess particular aspects of speech and language. For this reason, many have seen an opportunity to leverage signal processing and machine learning applications to objectively assess clinical speech samples in order to automatically compute objective measures of neurological health. This document summarizes several contributions to expand upon this body of research. Mainly, there is still a large gap between the theoretical power of computational language models and their actual use in clinical applications. One of the largest concerns is the limited and inconsistent reliability of speech and language features used in models for assessing specific aspects of mental health; numerous methods may exist to measure the same or similar constructs and lead researchers to different conclusions in different studies. To address this, a novel measurement model based on a theoretical framework of speech production is used to motivate feature selection, while also performing a smoothing operation on features across several domains of interest. Then, these composite features are used to perform a much wider range of analyses than is typical of previous studies, looking at everything from diagnosis to functional competency assessments. Lastly, potential improvements to address practical implementation challenges associated with the use of speech and language technology in a real-world environment are investigated. The goal of this work is to demonstrate the ability of speech and language technology to aid clinical practitioners toward improvements in quality of life outcomes for their patients.
ContributorsVoleti, Rohit Nihar Uttam (Author) / Berisha, Visar (Thesis advisor) / Liss, Julie M (Thesis advisor) / Turaga, Pavan (Committee member) / Spanias, Andreas (Committee member) / Arizona State University (Publisher)
Created2022
Description
Information processing in the brain is mediated by network interactions between anatomically distant (centimeters apart) regions of cortex and network action is fundamental to human behavior. Disruptive activity of these networks may allow a variety of diseases to develop. Degradation or loss of network function in the brain can affect many aspects of the human experience; motor disorder, language difficulties, memory loss, mood swings, and more. The cortico-basal ganglia loop is a system of networks in the brain between the cortex, basal ganglia, the thalamus, and back to the cortex. It is not one singular circuit, but rather a series of parallel circuits that are relevant towards motor output, motor planning, and motivation and reward. Studying the relationship between basal ganglia neurons and cortical local field potentials may lead to insights about neurodegenerative diseases and how these diseases change the cortico-basal ganglia circuit. Speech and language are uniquely human and require the coactivation of several brain regions. The various aspects of language are spread over the temporal lobe and parts of the occipital, parietal, and frontal lobe. However, the core network for speech production involves collaboration between phonologic retrieval (encoding ideas into syllabic representations) from Wernicke’s area, and phonemic encoding (translating syllables into motor articulations) from Broca’s area. Studying the coactivation of these brain regions during a repetitive speech production task may lead to a greater understanding of their electrophysiological functional connectivity. The primary purpose of the work presented in this document is to validate the use of subdural microelectrodes in electrophysiological functional connectivity research as these devices best match the spatial and temporal scales of brain activity. Neuron populations in the cortex are organized into functional units called cortical columns. These cortical columns operate on the sub-millisecond temporal and millimeter spatial scale. The study of brain networks, both in healthy and unwell individuals, may reveal new methodologies of treatment or management for disease and injury, as well as contribute to our scientific understanding of how the brain works.
ContributorsO'Neill, Kevin John (Author) / Greger, Bradley (Thesis advisor) / Santello, Marco (Committee member) / Helms Tillery, Stephen (Committee member) / Papandreou-Suppapola, Antonia (Committee member) / Kleim, Jeffery (Committee member) / Arizona State University (Publisher)
Created2021