Global ETD Search

Search theses and dissertations gathered from participating repositories worldwide. Every result links back to the library that holds it. No account is needed.

Results

Showing 1 to 12 of 12 for “"model pre-training"”.

  1. Prior knowledge-driven model pre-training and fine-tuning for ophthalmic image analysis

    … of conventional fine-tuning strategies for large pre-trained models (LPMs). This thesis addresses these challenges by developing and evaluating novel pre-training and fine-tuning frameworks tailored for ophthalmic imaging. We first establish a strong and reproducible baseline for ophthalmic …

    ubc Repository record for Prior knowledge-driven model pre-training and fine-tuning for ophthalmic image analysis (opens in a new tab) Download PDF for Prior knowledge-driven model pre-training and fine-tuning for ophthalmic image analysis (opens in a new tab)

  2. Weakly-supervised text classification

    … text classification task, due to their strong expressive power and less requirement for feature engineering. Despite such attractiveness, neural text classification models suffer from the lack of training data in many real-world applications. Although many semi-supervised and weakly-supervised …

    uiuc Repository record for Weakly-supervised text classification (opens in a new tab)

  3. PRACTICAL NAMED ENTITY RECOGNITION: THE ROLE OF ENTITY AND ITS CONTEXT

    Neural supervised models for named entity recognition perform well within the same domain but fail to recognize entities not seen in the (pre-)training data with high accuracy. For better generalization, it is essential that models are able to recognize predictive contextual clues. In this thesis, …

    penn Repository record for PRACTICAL NAMED ENTITY RECOGNITION: THE ROLE OF ENTITY AND ITS CONTEXT (opens in a new tab)

  4. Attention based contrastive learning with non-Siamese architecture

    Learning compressed representations of higher-dimensional data for exploratory analysis and discovery of semantically meaningful information is a long-standing problem in computer vision. Self-supervised methods such as contrastive learning have been employed for model pre-training with good …

    iastate Repository record for Attention based contrastive learning with non-Siamese architecture (opens in a new tab)

  5. Efficient Distributed Training With Subnetworks

    Pre-training large neural networks at scale places significant memory and communication demands on modern accelerators. As model sizes continue to grow faster than available device memory, efficiently distributing training across multiple hardware devices has become increasingly challenging. In …

    concordia Download PDF for Efficient Distributed Training With Subnetworks (opens in a new tab)

  6. Explainable AI in satellite scheduling : integrating large language models, knowledge graphs, and computational argumentation for onboard and ground segment systems

    … possible. Through integrating with systems and models, XAI techniques assist with validating or correcting system behaviour, which can substantiate trust and grant assurances that training and model configuration are functioning as designed. Various approaches for XAI can generate textual and …

    strathclyde Repository record for Explainable AI in satellite scheduling : integrating large language models, knowledge graphs, and computational argumentation for onboard and ground segment systems (opens in a new tab)

  7. AI-guided detection of antibiotic-resistantbacteria using resistance genes

    … vicious resistance cycle. Transformer-based AI models have shown state-of-the-art performance in understanding complex patterns in data. The thesis aimed to create a framework on how to implement transformers to predict bacterial resistance profiles by training on genomic data. The framework …

    chalmers Repository record for AI-guided detection of antibiotic-resistantbacteria using resistance genes (opens in a new tab)

  8. Turn of Phrase: Contrastive Pre-Training for Discourse-Aware Conversation Models

    … conversation. Recent advances in unsupervised representation learning of text have been attained primarily through language modeling, which models discourse only implicitly and within a small window. These representations are in turn evaluated chiefly on sentence pair or paragraph-question pair …

    byu Repository record for Turn of Phrase: Contrastive Pre-Training for Discourse-Aware Conversation Models (opens in a new tab) Download PDF for Turn of Phrase: Contrastive Pre-Training for Discourse-Aware Conversation Models (opens in a new tab)

  9. GASTON: Graph-Aware Social Transformer for Online Networks

    … Mitigating these harms requires computational models that can understand the nuance of online interactions to accurately detect harmful content such as toxicity and norm violation. This is difficult because the meaning of an individual post is rarely self-contained; it is dynamically …

    waterloo Repository record for GASTON: Graph-Aware Social Transformer for Online Networks (opens in a new tab)

  10. Applying Transformer-Based Deep Learning Model for Predicting Multimorbidity in Older Adults

    Disease predictive modelling supports decision-making for policymakers and healthcare providers, and aids in managing health conditions for individuals. Recent advances in deep learning have exhibited outstanding accuracy in diagnosis prediction. Bidirectional Encoder Representations from …

    sask Repository record for Applying Transformer-Based Deep Learning Model for Predicting Multimorbidity in Older Adults (opens in a new tab)

  11. USING CNNS TO UNDERSTAND LIGHTING WITHOUT REAL LABELED TRAINING DATA

    … of the lighting are impossible to achieve. As a preliminary topic, we rst study the capacity of deep CNNs. Designing deep CNNs with less capacity and good generalization is one way to reduce the amount of labeled data needed in training deep CNNs, and understanding the capacity of deep CNNs is …

    maryland Repository record for USING CNNS TO UNDERSTAND LIGHTING WITHOUT REAL LABELED TRAINING DATA (opens in a new tab)