{"id":{"repo_id":"denver","oai_identifier":"oai:digitalcommons.du.edu:etd-3924"},"canonical_url":"https://search.ndltd.org/etd/denver/oai:digitalcommons.du.edu:etd-3924","repository":{"repo_id":"denver","name":"University of Denver","base_url":"https://digitalcommons.du.edu/do/oai/"},"display":{"title":"Addressing the Problems of Data Variations, Quality, and Scarcity in Training Deep Neural Networks","abstract":"The performance of deep neural networks (DNNs) is strongly influenced by the characteristics and quality of the underlying datasets. This Ph.D. dissertation addresses three pervasive data challenges-imbalance, quality degradation, and scarcity-that commonly hinder the effectiveness of DNNs in computer vision (CV) and natural language processing (NLP) applications. Class imbalance remains one of the most frequent causes of degraded model generalization. While Focal Loss effectively mitigates inter-class imbalance by assigning higher weights to minority classes, it struggles with intra-class imbalance, particularly in video datasets where longer clips dominate feature representation. To address this, I implement and utilize the AD-CORRE Loss, which operates at the mini-batch level to balance intra-class feature contributions with minimal computational overhead. Combined with Focal Loss, this dual-loss framework yields more stable and robust training, as demonstrated by improved empirical results across multiple CV benchmarks. Data quality represents another major factor influencing learning outcomes. Blurred or low-resolution videos and images reduce feature richness and lead to biased feature extraction. To account for this, I introduce the Combined-SSL (Self-Supervised Learning) mechanism, which jointly models video quality and classification. The integration of quality-aware supervision significantly enhances recognition performance on challenging datasets. Moreover, in an NLP setting, I apply a RoBERTa-CNN model to detect suicide intentions from well-cleaned social media posts, further demonstrating the critical role of data quality in reliable model prediction. Data scarcity poses a fundamental limitation for supervised learning, especially when labeled samples are rare or sequences are short. To alleviate this, I design a multi-task learning framework that integrates auxiliary tasks-Masked Language Modeling (MLM) and S/TP prediction-within the ProtBERT backbone. Self-supervised and physics-informed tasks enrich the feature space and improve generalization. Experimental results on IDP datasets confirm that the proposed Multi-task ProtBERT effectively mitigates data scarcity and achieves state-of-the-art performance. Overall, this dissertation provides a unified investigation into data imbalance, data quality, and data scarcity-three core bottlenecks of modern deep learning-and proposes principled solutions that improve robustness, interpretability, and efficiency across both CV and NLP domains.","abstract_html":"The performance of deep neural networks (DNNs) is strongly influenced by the characteristics and quality of the underlying datasets. This Ph.D. dissertation addresses three pervasive data challenges-imbalance, quality degradation, and scarcity-that commonly hinder the effectiveness of DNNs in computer vision (CV) and natural language processing (NLP) applications. Class imbalance remains one of the most frequent causes of degraded model generalization. While Focal Loss effectively mitigates inter-class imbalance by assigning higher weights to minority classes, it struggles with intra-class imbalance, particularly in video datasets where longer clips dominate feature representation. To address this, I implement and utilize the AD-CORRE Loss, which operates at the mini-batch level to balance intra-class feature contributions with minimal computational overhead. Combined with Focal Loss, this dual-loss framework yields more stable and robust training, as demonstrated by improved empirical results across multiple CV benchmarks. Data quality represents another major factor influencing learning outcomes. Blurred or low-resolution videos and images reduce feature richness and lead to biased feature extraction. To account for this, I introduce the Combined-SSL (Self-Supervised Learning) mechanism, which jointly models video quality and classification. The integration of quality-aware supervision significantly enhances recognition performance on challenging datasets. Moreover, in an NLP setting, I apply a RoBERTa-CNN model to detect suicide intentions from well-cleaned social media posts, further demonstrating the critical role of data quality in reliable model prediction. Data scarcity poses a fundamental limitation for supervised learning, especially when labeled samples are rare or sequences are short. To alleviate this, I design a multi-task learning framework that integrates auxiliary tasks-Masked Language Modeling (MLM) and S/TP prediction-within the ProtBERT backbone. Self-supervised and physics-informed tasks enrich the feature space and improve generalization. Experimental results on IDP datasets confirm that the proposed Multi-task ProtBERT effectively mitigates data scarcity and achieves state-of-the-art performance. Overall, this dissertation provides a unified investigation into data imbalance, data quality, and data scarcity-three core bottlenecks of modern deep learning-and proposes principled solutions that improve robustness, interpretability, and efficiency across both CV and NLP domains.","abstract_has_math":false,"creators":["Sun, Jian"],"institution":null,"degree_name":"Ph.D. in Computer Science","degree_level":"Dissertation","degree_discipline":null,"degree_department":null,"school":null,"contributors":["Mohammad H. Mahoor","Kingshuk Ghosh","Haluk Ogmen","Matt J. Rutherford"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":null,"date_issued":"","date_published":null,"updated_at":"2026-11-19T07:55:18Z","subjects":["Computer vision","Data imbalance","Data quality","Data scarcity","Machine learning","Natural language processing (NLP)","Artificial Intelligence and Robotics","Computer Sciences","Data Science","Theory and Algorithms"],"languages":["English (eng)"],"rights":["<p>Copyright is held by the author. User is responsible for all copyright compliance.</p>","Jian Sun"],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier","label":"Identifier","values":["ETDCRP.2026.66","<a href=\"https://doi.org/10.56902/ETDCRP.2026.66\">https://doi.org/10.56902/ETDCRP.2026.66</a>","https://digitalcommons.du.edu/etd/2685"],"render_values":[{"text":"ETDCRP.2026.66","href":null,"code":true},{"text":"<a href=\"https://doi.org/10.56902/ETDCRP.2026.66\">https://doi.org/10.56902/ETDCRP.2026.66</a>","href":"https://doi.org/10.56902/ETDCRP.2026.66\">https://doi.org/10.56902/ETDCRP.2026.66</a>","code":true},{"text":"https://digitalcommons.du.edu/etd/2685","href":"https://digitalcommons.du.edu/etd/2685","code":true}]}],"person_identifier_entries":[]},"links":{"outbound_url":"https://doi.org/10.56902/ETDCRP.2026.66","outbound_label":"DOI","outbound_source":"dc:identifier","record_url":"https://doi.org/10.56902/ETDCRP.2026.66","record_label":"Repository record","record_kind":"doi","record_source":"dc:identifier","download_url":null,"download_label":null,"download_source":null,"access_status":null,"embargo_until":null,"access_note":null},"source_record":{"url":"https://digitalcommons.du.edu/do/oai/?verb=GetRecord&metadataPrefix=oai_etdms&identifier=oai%3Adigitalcommons.du.edu%3Aetd-3924","prefix":"oai_etdms"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Mohammad H. Mahoor","Kingshuk Ghosh","Haluk Ogmen","Matt J. Rutherford"]},{"key":"dc:creator","label":"Author","values":["Sun, Jian"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.copyright","label":"Dc Date Copyright","values":["2026-03-01T08:00:00Z"]},{"key":"dc:date.created","label":"Dc Date Created","values":["2026-03-21T07:00:00Z"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Dissertation"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Ph.D. in Computer Science"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Computer vision","Data imbalance","Data quality","Data scarcity","Machine learning","Natural language processing (NLP)","Artificial Intelligence and Robotics","Computer Sciences","Data Science","Theory and Algorithms"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["English (eng)"]},{"key":"dc:rights","label":"Dc Rights","values":["<p>Copyright is held by the author. User is responsible for all copyright compliance.</p>"]},{"key":"dc:rights.holder","label":"Dc Rights Holder","values":["Jian Sun"]},{"key":"dc:rights.license","label":"Dc Rights License","values":["all_rights_reserved"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["ETDCRP.2026.66","https://doi.org/10.56902/ETDCRP.2026.66","<a href=\"https://doi.org/10.56902/ETDCRP.2026.66\">https://doi.org/10.56902/ETDCRP.2026.66</a>","https://digitalcommons.du.edu/etd/2685"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["The performance of deep neural networks (DNNs) is strongly influenced by the characteristics and quality of the underlying datasets. This Ph.D. dissertation addresses three pervasive data challenges-imbalance, quality degradation, and scarcity-that commonly hinder the effectiveness of DNNs in computer vision (CV) and natural language processing (NLP) applications. Class imbalance remains one of the most frequent causes of degraded model generalization. While Focal Loss effectively mitigates inter-class imbalance by assigning higher weights to minority classes, it struggles with intra-class imbalance, particularly in video datasets where longer clips dominate feature representation. To address this, I implement and utilize the AD-CORRE Loss, which operates at the mini-batch level to balance intra-class feature contributions with minimal computational overhead. Combined with Focal Loss, this dual-loss framework yields more stable and robust training, as demonstrated by improved empirical results across multiple CV benchmarks. Data quality represents another major factor influencing learning outcomes. Blurred or low-resolution videos and images reduce feature richness and lead to biased feature extraction. To account for this, I introduce the Combined-SSL (Self-Supervised Learning) mechanism, which jointly models video quality and classification. The integration of quality-aware supervision significantly enhances recognition performance on challenging datasets. Moreover, in an NLP setting, I apply a RoBERTa-CNN model to detect suicide intentions from well-cleaned social media posts, further demonstrating the critical role of data quality in reliable model prediction. Data scarcity poses a fundamental limitation for supervised learning, especially when labeled samples are rare or sequences are short. To alleviate this, I design a multi-task learning framework that integrates auxiliary tasks-Masked Language Modeling (MLM) and S/TP prediction-within the ProtBERT backbone. Self-supervised and physics-informed tasks enrich the feature space and improve generalization. Experimental results on IDP datasets confirm that the proposed Multi-task ProtBERT effectively mitigates data scarcity and achieves state-of-the-art performance. Overall, this dissertation provides a unified investigation into data imbalance, data quality, and data scarcity-three core bottlenecks of modern deep learning-and proposes principled solutions that improve robustness, interpretability, and efficiency across both CV and NLP domains."]},{"key":"dc:description.provenance","label":"Dc Description Provenance","values":["Received from ProQuest"]},{"key":"dc:format","label":"Dc Format","values":["application/pdf"]},{"key":"dc:format.extent","label":"Dc Format Extent","values":["183 pgs","14.2 MB"]},{"key":"dc:title","label":"Title","values":["Addressing the Problems of Data Variations, Quality, and Scarcity in Training Deep Neural Networks"]}]}],"canonical_facts":{"dc:contributor":["Mohammad H. Mahoor","Kingshuk Ghosh","Haluk Ogmen","Matt J. Rutherford"],"dc:creator":["Sun, Jian"],"dc:date.copyright":["2026-03-01T08:00:00Z"],"dc:date.created":["2026-03-21T07:00:00Z"],"dc:description.abstract":["The performance of deep neural networks (DNNs) is strongly influenced by the characteristics and quality of the underlying datasets. This Ph.D. dissertation addresses three pervasive data challenges-imbalance, quality degradation, and scarcity-that commonly hinder the effectiveness of DNNs in computer vision (CV) and natural language processing (NLP) applications. Class imbalance remains one of the most frequent causes of degraded model generalization. While Focal Loss effectively mitigates inter-class imbalance by assigning higher weights to minority classes, it struggles with intra-class imbalance, particularly in video datasets where longer clips dominate feature representation. To address this, I implement and utilize the AD-CORRE Loss, which operates at the mini-batch level to balance intra-class feature contributions with minimal computational overhead. Combined with Focal Loss, this dual-loss framework yields more stable and robust training, as demonstrated by improved empirical results across multiple CV benchmarks. Data quality represents another major factor influencing learning outcomes. Blurred or low-resolution videos and images reduce feature richness and lead to biased feature extraction. To account for this, I introduce the Combined-SSL (Self-Supervised Learning) mechanism, which jointly models video quality and classification. The integration of quality-aware supervision significantly enhances recognition performance on challenging datasets. Moreover, in an NLP setting, I apply a RoBERTa-CNN model to detect suicide intentions from well-cleaned social media posts, further demonstrating the critical role of data quality in reliable model prediction. Data scarcity poses a fundamental limitation for supervised learning, especially when labeled samples are rare or sequences are short. To alleviate this, I design a multi-task learning framework that integrates auxiliary tasks-Masked Language Modeling (MLM) and S/TP prediction-within the ProtBERT backbone. Self-supervised and physics-informed tasks enrich the feature space and improve generalization. Experimental results on IDP datasets confirm that the proposed Multi-task ProtBERT effectively mitigates data scarcity and achieves state-of-the-art performance. Overall, this dissertation provides a unified investigation into data imbalance, data quality, and data scarcity-three core bottlenecks of modern deep learning-and proposes principled solutions that improve robustness, interpretability, and efficiency across both CV and NLP domains."],"dc:description.provenance":["Received from ProQuest"],"dc:format":["application/pdf"],"dc:format.extent":["183 pgs","14.2 MB"],"dc:identifier":["ETDCRP.2026.66","https://doi.org/10.56902/ETDCRP.2026.66","<a href=\"https://doi.org/10.56902/ETDCRP.2026.66\">https://doi.org/10.56902/ETDCRP.2026.66</a>","https://digitalcommons.du.edu/etd/2685"],"dc:language":["English (eng)"],"dc:rights":["<p>Copyright is held by the author. User is responsible for all copyright compliance.</p>"],"dc:rights.holder":["Jian Sun"],"dc:rights.license":["all_rights_reserved"],"dc:subject":["Computer vision","Data imbalance","Data quality","Data scarcity","Machine learning","Natural language processing (NLP)","Artificial Intelligence and Robotics","Computer Sciences","Data Science","Theory and Algorithms"],"dc:title":["Addressing the Problems of Data Variations, Quality, and Scarcity in Training Deep Neural Networks"],"thesis:degree_level":["Dissertation"],"thesis:degree_name":["Ph.D. in Computer Science"]},"updated_at":"2026-11-19T07:55:18Z"}