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 27 for “"Neural ODE"”.
-
Can a Neural ODE learn a Chaotic System?
… to initial conditions. Conventional metrics of model performance, such as generalization error, often fail to capture a neural network's ability to reproduce the invariant statistics of these complex systems. In this thesis, we demonstrate through a comprehensive set of examples that pointwise …
-
Designing and Evaluating Neural Ordinary Differential Equation Models for Pharmacokinetic and Pharmacodynamic Time Series
PKPD modeling is the study of the effect of drugs on living organisms. It combines modeling the concentration (PK) and effect (PD) of a drug over time. This is commonly done through systems of ordinary differential equations (ODEs), although there have been increasing efforts to utilize machine …
-
Data Driven Approaches to Predict and Control Chinese Hamster Ovary Cell Growth in a Bioreactor System
… process stability pose significant challenges. Modeling methods that capture the dynamics of variables and process parameters in cell culture significantly enhance manufacturing efficiency. Meanwhile, automated process controllers stabilize operations and mitigate batch-to-batch variations. …
-
Toward Utilizing Analog Rydberg Atom Systems for Quantum Machine Learning
… computers for discriminative tasks, specifically neural ordinary differential equation (neural ODE) based residual neural network (ResNets) classifiers. ResQ demonstrates how analog quantum computing can be tailored to solve classification problems in machine learning, achieving a 36% improvement …
-
MULTI-SCALE DATA-EFFICIENT MACHINE LEARNING FRAMEWORK FOR ACCELERATED DISCOVERY OF HETEROGENEOUS CATALYSTS
… At nanoparticle scale, a pre-trained foundation model then predicts nanoparticle surface exposure and synthesizability at DFT-level accuracy. At reactor scale, a domain-knowledge-informed neural ODE model provides interpretable and data-efficient dynamic predictions for flow reactors without …
-
Deep learning based control for power converters
… the increasing complexity and dynamic nature of modern power systems, there has been a growing interest in applying data-driven techniques—particularly deep learning—to enhance control performance, improve system resilience, and reduce the need for explicit mathematical modeling. In this context, …
-
Asymptotic analysis of deep learning algorithms
… different from those implicitly assumed in the neural ODE literature. We study the convergence of the hidden state dynamics in these scaling regimes, showing that one may obtain an ODE, a stochastic differential equation (SDE) or neither. Furthermore, we derive the corresponding scaling limits …
-
Boundary-Aware Neural Point-to-Point Set Evolution for Interactive Trajectory Prediction
… prediction in multi-agent settings requires models that can handle complex interactions under real-world sensing constraints, such as partial observations and temporal irregularities. This thesis presents a unified framework for interactive trajectory prediction that is grounded in deep …
-
Machine learning applications in astrophysics: Reduced-order modelling for chemical kinetics and galaxy merger reconstruction with graph neural network
… chemistry in simulations. The combination of neural ordinary differential equations and autoencoders is shown to be a promising approach for reducing the complexity of chemical kinetics simulations. Specifically, autoencoders identify the reduced reaction subspace while the neural ODE learn …
-
Explainable Artificial Intelligence Techniques in Medical Imaging Analysis
… the image feature through hierarchical data modeling directly from the input data. Both classic ML and DL models have emerged as promising AI tools for medical image analysis. Despite promising academic research in which algorithms are beginning to outperform humans, clinical radiography …
-
Modelling Continuous-Time Earth System Dynamics for Data-Driven Forecasting, Emulation, and Downscaling
… Weather Prediction (NWP) and Earth System Models (ESMs) provide strong predictive skill but rely on computationally expensive, high-resolution simulations. CMIP6-class climate models require thousands of CPU-hours per simulated year, while operational \(0.25^{\circ}\) forecasts involve dense …
-
Hierarchical Planning for Safe and Reactive Robot Autonomy in Dynamic, Human-centric Environments
… domain of interactive manipulation, we augment Neural ODE models with online correction terms for stability and safety, and integrate a reactive temporal-logic task planner, thereby enabling a Franka robot arm to seamlessly adapt to disturbances and human interactions for wiping and stirring …
-
Insights into neural network learning of structured models : optimization, architectures, and scientific surrogates
… this thesis investigates the interplay between neural network architecture, loss landscape geometry, and optimization in scientific machine learning. We analyze optimization of physics-informed neural networks (PINNs), which can exhibit ill-conditioned loss landscapes. We develop a randomized …
-
Statistical Deep Learning of Multivariate Longitudinal Data
… longitudinal data presents significant modeling challenges. For years, scientific research has been actively exploring dynamic interactions among multiple components and understanding how interventions can impact outcomes over time with complex underlying dynamics. However, statistical …
-
Computational modeling of the injection molding process: open-source methods and data-driven fiber orientation
… molding (IM) cycle. Despite progress in IM modeling, there remains a clear need for transparent, reproducible and verifiable numerical tools that are easily extensible and adapt to application-specific requirements. Many of the existing solutions are proprietary (closed-source) and difficult …
-
UNDERSTANDING WHOLE BRAIN ACTIVITY THROUGH BRAIN NETWORK MODELS
The central nervous system coordinates many neural subpopulations connected via macroscale white matter architecture and surface cortical connections to produce complex behavior depending on environmental cues. The activity occurs over different scales, from information transfer between individual …
-
Learning from What’s Missing: Masked Approaches for Robust and Interpretable Deep Models
… to the spatial content of each image patch. Deep neural networks trained under self-supervised paradigms rely excessively on narrow local cues and struggle with incomplete data or large domain shifts. This dissertation explores partial-visibility strategies in three contrasting application areas: …
-
Differential Equations for Machine Learning
Mechanistic models express novel hypotheses for an observed phenomenon by constructing mathematical formulations of causal mechanisms. As opposed to this modeling paradigm, machine learning approaches learn input-output mappings by complicated and often non-interpretable models. While requiring …
-
Graph-based Multi-ODE Neural Networks for Spatio-Temporal Traffic Forecasting
… the development of spatio-temporal forecasting models in many applications, and traffic forecasting is one of the most important ones. Long-range traffic forecasting, however, remains a challenging task due to the intricate and extensive spatio-temporal correlations observed in traffic networks. …
-
Neural network enhanced off-road skid-steer vehicle modeling with an application to path planning
Submission published under a 24 month embargo labeled 'Closed Access', the embargo will last until 2024-08-01
Page 1 of 2