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 20 of 30 for “"parameter efficiency"”.
-
On Parameter Efficiency of Neural Language Models
… of scaling these models to encompass billions of parameters, while enhancing adaptability and emergent capabilities, has brought forth significant deployment challenges due to their massive size. These challenges include constraints in model storage and inference latency for real-world deployment, …
-
Robust and Efficient Learning for Visual Perception
… novel architectures that optimize training parameters and inference latency, this work ensures high-performance perception under strict resource constraints. Another key contribution is the proposal of large-scale benchmarks for both tasks, designed to rigorously evaluate model efficiency …
-
TOWARDS EFFICIENT TRANSFORMER SCALING
… explores strategies to enhance Transformer efficiency. First, we propose WideNet, which optimizes parameter efficiency using parameter-sharing and Mixture-of-Experts, achieving strong results in both vision and language tasks. Second, we investigate transformer configurations, finding that …
-
Semantic communication for video coding and transmission
… by methods for Deep Neural Network (DNN) parameter efficiency. For hybrid coding, the thesis introduces a hybrid semantic–VVC video codec, extends it with hierarchical bidirectional frame prediction, and further develops it into a context conditioned architecture based on a shared …
-
Boundary-Aware Neural Point-to-Point Set Evolution for Interactive Trajectory Prediction
… in JAX, is a VAE that demonstrates significant parameter efficiency. Experimental results on standard benchmarks show that our models are competitive with state-of-the-art methods. Furthermore, we demonstrate that our final architecture successfully handles the more challenging task of …
-
Achieving More with Less: Learning Generalizable Neural Networks With Less Labeled Data and Computational Overheads
… devices. This dissertation aims to improve efficiency in machine learning by exploring different ways to learn generalizable neural networks that require less labeled data and computational resources. We demonstrate that using physics supervision in scientific problems can reduce the need …
-
Efficient inference and adaptation in deep learning : from edge systems to large language models
… Consequently, improving the resource efficiency of DL throughout its lifecycle has become an increasingly important research problem.This thesis investigates efficient inference, training, and adaptation in DL across the model lifecycle from four complementary perspectives: (1) …
-
Interpretable Deep Image Denoiser by Unrolling Graph Laplacian Regularizer
… but these models require on large numbers of parameters, lack interpretability, and depend heavily on random parameter initialization. As a result, they frequently converge to poor-performing local minima. This thesis proposes an image denoising neural net constructed by unrolling an iterative …
-
Fine-tuning ASR Models for Very Low-Resource Languages: A Study on Mvskoke
… to evaluate performance in different settings. Parameter efficiency of adapter training is compared with training entire models, as well as examining the impact of the number of languages used during pre-training. The study also investigates how performance varies with different amounts of …
-
Controlling Content Selection in Biomedical Summarization and Question Answering
… while training under one percent of the backbone parameters and matching full fine-tuning on summarization quality. Across both loci, structure substitutes for scale: external selection yields traceability and modularity, internal selection parameter efficiency at the cost of observability.
-
Superpixel Salient Object Detection
… Although superpixels offer significant efficiency advantages, their irregular and heterogeneous nature presents unique challenges for integration into modern deep learning frameworks. Furthermore, traditional augmentation strategies and transfer learning pipelines are not readily …
-
DATA-DRIVEN AND CONSTRAINED APPROACHES TO LEARNING AND ENFORCING SYMMETRY IN NEURAL NETWORKS
… jointly learning augmentation policies and model parameters, and use PAC-Bayesian theory to derive conditions under which the learned distribution recovers unknown invariances. For equivariant architectures, we first demonstrate their necessity through 3D reconstruction and registration tasks, …
-
On the Resource Efficiency of Language Models
… resource challenges in two dimensions: data efficiency and model efficiency. For post-training, LLMs face data curation challenges where high-quality labeled data is scarce and expensive to obtain, and data utilization challenges where existing methods fail to optimize model performance on …
-
Robust Audiovisual Active Speaker Detection for Egocentric Recordings - Leveraging Contextual Information and Cross-Modal Biometric Associations
… under egocentric conditions while preserving parameter efficiency. The resulting models achieve competitive or superior performance to heavier synchronisation-based systems on large-scale benchmarks. More broadly, the findings highlight the potential of context-enriched multimodal …
-
Parameter-Efficient Fine-Tuning for Learning Language Model Watermarks
… models. This thesis explores two different parameter-efficient fine-tuning methods for training large language models to mimic watermarked models. These methods are: (i) the low rank adaptations and (ii) transfer learning. The preliminary results show that a relatively small network can …
-
Parameter-Efficient Fine-Tuning for Learning Language Model Watermarks
… models. This thesis explores two different parameter-efficient fine-tuning methods for training large language models to mimic watermarked models. These methods are: (i) the low rank adaptations and (ii) transfer learning. The preliminary results show that a relatively small network can …
-
Parameter-Efficient Fine-Tuning for Learning Language Model Watermarks
… models. This thesis explores two different parameter-efficient fine-tuning methods for training large language models to mimic watermarked models. These methods are: (i) the low rank adaptations and (ii) transfer learning. The preliminary results show that a relatively small network can …
-
Parameter-Efficient Fine-Tuning for Learning Language Model Watermarks
… models. This thesis explores two different parameter-efficient fine-tuning methods for training large language models to mimic watermarked models. These methods are: (i) the low rank adaptations and (ii) transfer learning. The preliminary results show that a relatively small network can …
-
APPLICATIONS OF DOMAIN DIVERGENCES FOR DOMAIN ADAPTATION IN NLP
… a method for applying divergence measures in a parameter-efficient manner for domain adaptation in NLP. The work concludes with limitations and future research avenues.
-
Capability-Aware Shared Hypernetworks for Heterogeneous Multi-Agent Coordination
… span a spectrum between two endpoints: i) shared-parameter methods, which assign an ID to each agent to encode diverse behaviors within a single architecture for sample-efficiency, but are limited in their ability to learn diverse behaviors; ii) independent methods, which learn a separate policy …
Page 1 of 2