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 20 for “"Deep learning theory"”.
-
Subsurface characterization using deep learning theory and techniques
… available measurements. Additionally, a deep learning surrogate model is used for faster simulations. Chapter 3 applies the proposed data assimilation framework to CO2 storage sites characterization. Several subsurface site characterization examples including Gaussian, channelized, and …
-
Nonlinear and geometric control methods in deep learning theory
Submission published under a 24 month embargo labeled 'U of I Access', the embargo will last until 2026-12-01
-
Demystifying deep network architectures : from theory to applications
Deep neural networks significantly power the success of machine learning and artificial intelligence. Over the past decade, the community keeps designing architectures of deep layers and complicated connections. Many works in deep learning theory tried to understand deep networks from different …
-
Exploring the relationship between Deep Neural Networks and Neural Tangent Kernel via covariance manipulation
… this study, we examine the relationship between Deep Neural Networks (DNNs) and the Neural Tangent Kernel (NTK), with a particular emphasis on covariance manipulation. Using the foundational framework provided by Roberts and Yaida in their book, "Deep Learning Theory," we methodically explore the …
-
Towards Newtonian understanding of deep learning training dynamics
Ever since the deep learning revolution, neural networks have been the dominating models in different machine learning areas such as computer vision, natural language processing, speech recognition. The wide application and success of deep learning in different areas have also lead to the evolution …
-
Machine Learning As Tool And Theory For Computational Neuroscience
… brain based on the ideas and methods of machine learning. This is effort has been encouraged, in part, by recent advances in neural network models. It is also driven by a recognition of the complexity of neural computation and the challenges that this poses for neuroscience’s methods. In this …
-
Testing theories of computation and learning in the visual cortex
… coding. In the second study, we propose that learning dynamics – the evolving patterns of neural changes across time compared within and across areas – can serve as a crucial hallmark to test theories of learning in the brain. We devise a multi-stage behaviour and neuronal imaging paradigm in …
-
Understanding and Enhancing Machine Learning Models with Theoretical Foundations
Machine learning has become a key driver of many contemporary technological advancements. With its empirical success, there is an urgent need for theoretical research to explain and complement these practical achievements. This includes understanding the empirical success of machine learning, …
-
Deep learning frameworks for modeling how neural circuits learn
The brain's prowess in learning and adapting remains an enigma, particularly in its approach to the 'temporal credit assignment' problem. How do neural circuits determine which specific states and connections contribute to future outcomes, and subsequently adjust these for enhanced learning? My …
-
Learning with Distribution Shift: Theory and Practice
… design and theoretical analysis in continual learning (CL), a machine learning paradigm where learning agents sequentially encounter tasks with distinct distributions. While state-of-the-art machine learning algorithms (e.g., deep neural networks) designed for single-task problems achieve …
-
Understanding Modern Machine Learning Models through the Llens of High-dimensional Statistics
Deep neural networks have become an indispensable component of artificial intelligence as seen in a wide range of impressive applications, yet many of their properties remain mysterious. For instance, the non-convex loss landscape may conflict with efficient optimization by standard gradient-based …
-
Analysis of Learning Algorithms via Diffusion Limits
Modern machine learning algorithms are often very difficult to analyze, as discrete units of scaling and time typically do not lead to very clean closed form representations. For example, updates can be in discrete time, and neural networks have finitely many distinct neurons per layer. In this …
-
Uniform Sampling from Deep Generative Network Manifolds
Deep Generative Networks (DGNs) are extensively employed to approximate data manifolds and distributions. However, samples used to train such DGNs are often distributed non-uniformly on the manifold, due to the cost or convenience of collection. For example, the CelebA dataset contains a large …
-
Non-convex and Interactive Learning via Stochastic Optimization
Advances in machine learning have led to many empirical breakthroughs in computer science, from developing state-of-the-art image classification models to defeating the world champion in the game of Go. From statistical learning (e.g. image recognition) to interactive learning (e.g. the game of …
-
STATISTICAL AND HIGH-DIMENSIONAL PERSPECTIVES ON MACHINE LEARNING
… calibration.While the accuracy of modern machine learning techniques continues to improve, many models exhibit mis-calibration, wherein the probability scores produced by the models fail to align with the actual frequencies of the labels. This discrepancy can lead to unreliable predictions and …
-
STRUCTURED DATA ANALYSIS: MODELS, ALGORITHMS AND THEORIES
… problems in computer vision and machine learning becomes very important. This thesis addresses this problem from three aspects: (1) designing effective structured data learning models, (2) developing efficient optimization algorithms, and (3) establishing solid theory to understand …
-
Understanding Deep Learning via Tractable Models
Deep learning systems have transformed information processing, achieving remarkable success in domains ranging from computer vision to natural language processing, and are completely reshaping every sector of the industry. Yet, our theoretical understanding of why these systems work lags far behind …
-
The Angle Degeneracy Phenomenon in Deep Neural Networks: Analysis and Relation to Training Dynamics
Deep neural networks have proven to be powerful functions with many applications, but the theoretical behaviour of these functions is not fully understood. One such behaviour is the large depth degeneracy phenomenon, where inputs tend to become highly correlated as they travel deeper into a …