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 17 of 17 for “"Gradient based learning"”.
-
IMPROVED GRADIENT-BASED LEARNING FOR ENHANCED EFFICIENCY AND GENERALIZATION OF DEEP NEURAL NETWORKS
… the efficiency and generalization ability of gradient-based learning in deep neural networks (DNNs). The research begins by examining Sharpness-Aware Minimization (SAM) and proposes the Efficient Sharpness-Aware Minimizer (ESAM) as an improvement. ESAM reduces training costs by up to 30% while …
-
Theory and Applications of Gradient Free Optimization in Physics
Machine Learning (ML) has become a popular field of research in many domains. It has become a flexible option to tackle a large variety of problems. This Thesis examines a fundamental component of ML training to explore how these tools can be further used in physics. The gained knowledge is then …
-
Machine Learning with Differentiable Physics Priors
Differentiable physics priors enable gradient-based learning systems to adhere to physical dynamics. By making physics simulations differentiable, we can backpropagate through the physical consequences of actions. This pipeline allows agents to quickly learn to achieve desired effects in the …
-
Adaptive inverse modeling of a shape memory alloy wire actuator and tracking control with the model
… the weight parameters in both inverse models, a gradient-based learning algorithm is used. Finally, for the SMA position control, PID controller, adaptive controllers with KP model and adaptive nonlinear inverse model controller are compared experimentally.
-
Autonomous car driving using reinforcement learning and genetic algorithm /
This study explores applications of three learning strategies for autonomous racing in the TORCS simulation environment - NEAT (NeuroEvolution of Augmenting Topologies), DDPG (Deep Deterministic Policy Gradient), and Evolutionary Reinforcement Learning (ERL) – a hybrid algorithm created during this …
-
Evidence for Robustness Advantages of Neuroevolution Over Deep Reinforcement Learning
… problem in the field of Deep Reinforcement Learning (DRL). Previous works showed that small changes (often imperceptible to humans) in the testing environment such as perturbations, noise, adversarial agents, and training instabilities can cause catastrophic drop in the performance of DRL …
-
Deep Learning on Graphs: Theory, Models, Algorithms and Applications
… graphs are pervasive in our lives. Although deep learning has achieved tremendous success in many engineering fields, it is still limited in handling various structured data. More importantly, humans have the remarkable ability to learn discrete structures from data to facilitate explainability …
-
Intelligent Combination of Structural Analysis Algorithms: Application to Mathematical Expression Recognition
… analysis is an important step in many document based recognition problem. Structural analysis is performed to associate elements in a document and assign meaning to their association. Handwritten mathematical expression recognition is one such problem which has been studied and researched for …
-
Optimization of Learning Algorithms in Neuromorphic Computing Systems.
… which makes it difficult to apply traditional gradient-based learning techniques. Different approaches have been proposed to address this challenge, ranging from supervised learning - largely inspired by error backpropagation in Deep Neural Networks - to unsupervised learning, which closely …
-
Learning to Attend with Neural Networks
… available, artificial intelligence and machine learning researchers design ever larger and more complicated neural networks to learn from millions of data points. Although the traditional convolutional neural networks (CNNs) can achieve superhuman accuracy in object recognition tasks, they …
-
Exploring Patterns of Algorithmic Generalization in Deep Learning
… time and compute resources. However, deep learning systems have traditionally struggled with such flexibility. The typical neural network, such as a shallow MLP with a soft-thresholding non-linearity, processes fixed-sized inputs and outputs, consuming a constant amount of compute per …
-
Analog Alchemy: Neural Computation with In-Memory Inference, Learning and Routing
… dynamics of the devices are used for inference, learning and routing. Guided by the principles of gradient-based learning, we selected functions that need to be materialized, and analyzed connectomics principles for efficient wiring. Despite non-idealities and noise inherent in analog physics, I …
-
Learning and Decision-Making in Competitive and Uncertain Systems
… progressively more is being asked of machine learning methods. This evolution has lead to a renewed focus on learning and decision-making systems. In this domain, theoretical challenges relating to competition and uncertainty are emerging from the practical considerations that have motivated …
-
A Reinforcement Learning Guided Hybrid Neural Framework for Sparse Feature Selection
… Although deep neural networks and reinforcement learning (RL) techniques have shown promise, most existing methods remain static or semi-static, limiting their adaptability to nonstationary data. The core challenge, therefore, is to design a framework that can adaptively iden- tify sparse, …
-
Symmetry Induction in Computational Intelligence
… or whether changing reference to an intuition-based definition of symmetry over the evaluation function can also be of use. Within the context of optimization, symmetry induction as defined in this thesis will have the effect of equating the evaluation of a set of given objects. Group theory is …
-
Towards Precise Localization in Object Detection: Stable Matching Strategies and Advanced Bounding Box Regression Loss Functions in Deep Neural Networks
… alignment with ground truth boundaries. In gradient-based learning, these loss functions guide the optimization process, enhancing spatial alignment and improving detection accuracy. However, existing loss functions exhibit several limitations, such as sensitivity to scale variations, …