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 18 of 18 for “"Training Convergence"”.
-
ARCHITECTURE OPTIMIZATION, TRAINING CONVERGENCE AND NETWORK ESTIMATION ROBUSTNESS OF A FULLY CONNECTED RECURRENT NEURAL NETWORK
… Kalman filter (EKF) is adopted as a network's training algorithm. These two algorithms can seamlessly work together to generate the optimized RNN. The enhancement of the modeling performance of the optimized network comes from three parts: 1) its prototype - the FFCNN has advantages over …
-
Communication Design for Dynamic Distributed Learning Systems
… parallelized computation required for model training.Despite its advantages, distributed learning relies on frequent signal exchange and communication between the server and the data-holding entities throughout the learning process, making \textit{communication} a critical factor. Therefore, …
-
Reinforcement learning tuning of sliding mode control for an underactuated autonomous surface vehicle
… (PPO) algorithms are both used to compare the training of RL agents to estimate the required controller parameters. The system is implemented and evaluated in a MATLAB/Simulink environment using the Reinforcement Learning Toolbox. Simulation results were evaluated against a baseline data set …
-
REPRESENTATION LEARNING OF DATA WITH MULTIPLE MODALITIES WITH APPLICATIONS TO VISUAL QUESTION ANSWERING
… we propose a novel loss function to improve the training convergence and overall performance of complex, modular deep neural models of multimodal data.
-
Embedded artificial neural networks optimized for low-cost and low-size-memory devices
… in two different scenarios, embedded training, and inference. Analysis of training convergence, inference time duration, and program code occupation into the internal memories and a technique to optimize this implementation to fit bigger MLP architectures. The aim of this work is to …
-
Instance segmentation and material classification in X-ray computed tomography
… However, they require large amount of labeled training images, which are hard to obtain in CT research. This thesis develops a non-proposal 3-d instance segmentation and classification structure based on smoothed fully convolutional networks (FCNs), graph-based spatial clustering and ensembling …
-
AI - Driven Home Energy Management System for Profit and Grid Stability - Deep Reinforcement Learning and Predictive Models for Minimizing Peak Demand While Balancing Battery Degradation in a Dynamic Environment
… Although additional computational time for full training convergence and hyperparameter tuning remains as future work, these preliminary results underscore the strong potential of AI-driven residential energy management for cost savings and grid support.
-
Augmented Neural Network Surrogate Models for Polynomial Chaos Expansions and Reduced Order Modeling
… where the decisions made in the training process are not fully understood, and the roles of individual neurons are not well defined. We present two solutions towards this lack of interpretability. The first focuses on mimicking polynomial chaos (PC) modeling techniques, modifying …
-
Foundations of Diffusion Language Models
… transfer efficient parameterizations, improved training objectives, and distillation-based fast samplers to the discrete domain. Leveraging this, we obtain a twofold speedup in training convergence and two orders of magnitude faster sampling, and we introduce the first few-step discrete …
-
Accelerating Deep Learning by Optimising Communication and Computation Efficiency
… This thesis presents new optimisations for DNN training across multiple distributed machines and on a single machine. Our techniques reduce DNN training times, making DNNs faster and more efficient. First, we improve the communication efficiency of distributed training through multi-streamed …
-
Towards High Fidelity Generation of Synthetic 3D Worlds
… develops an important component in the full training pipeline: evaluation metrics, constructing datasets, and model architecture. In our first article we introduce a new metric for evaluating conditional generative adversarial networks (cGANs). Most standard metrics used for evaluating cGANs …
-
Mastering infinite depths: Optimization and generalization in deeper neural networks
… a significantly larger number of parameters than training samples, overparameterized neural networks have been widely adopted in practical applications, showcasing notable numerical and practical success. This success has sparked theoretical investigations aimed at understanding the underlying …
-
Learning-directed systems with safety and robustness certificates
… performance of learning-driven systems at training convergence. The framework tackles real-world challenges by combining formal verification with learning-based methods, enabling analysis of systems with neural network components, and adapting core algorithms to practical settings, such as …
-
Improving Pattern Recognition and Neural Network Algorithms With Applications to Solar Panel Energy Optimization
… improve pattern recognition and neural network training, initially motivated by a multidisciplinary effort to build a robot that helps maintain and optimize solar panel energy production. Our contributions detail an improved non-linear pre-processing technique to enhance poorly illuminated …
-
Network Resource Allocation and Topology Design for Distributed Machine Learning
… resources. The goal is to minimize overall training time across different learning paradigms, including server-based and decentralized peer-to-peer learning, from offline to online solutions. We first consider server-based distributed learning, where the server updates models by aggregating …
-
Perception and Action: From Sensorimotor Learning to Intelligent Tool-Use Behavior
… results show a higher success rate and faster training convergence time when proprioceptive and kinesthetic sensory feedback is used in addition to the visual input. Finally, on the last question, I studied the ability of a sensorimotor learning agent to use multiple types of tools. To achieve …
-
Deep Learning for 2D and 3D Scene Understanding
… objects effectively. A common issue in training deep neural networks is the need for large training sets. One approach to this is to use simulated image and video data, but this suffers from a domain gap wherein the performance on real-world data is poor relative to performance on the …
-
Neuronové sítě, předtrénování komplexity
Tato práce se zabývá technikou obrácené destilace modelu neuronových sítí, při které menší model (učitel) vede počáteční fázi tréninku většího modelu (studenta). Cílem je ověřit, zda tento přístup může sloužit jako efektivní způsob inicializace vah a zlepšit průběh tréninku. Experimenty na …