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 62 for “"Natural Gradient"”.
-
Computationally efficient natural gradient descent
… aim of creating a more computationally efficient natural gradient descent algorithm for on-line learning. It uses the statistical mechanics framework created by Saad and Solla to describe the evolution of order parameters for this algorithm in a two-layer student-teacher scenario, and compares …
-
Neural Networks and the Natural Gradient
… from the problem of local minima. The advent of natural gradient algorithms promised to overcome this shortcoming by finding better local minima. However, they require additional training parameters and computational overhead. By using a new formulation for the natural gradient, an algorithm is …
-
Natural gradient algorithms for training deep neural networks
… inspired artificial intelligence models). Natural Gradient Descent (NGD) algorithms have favorable properties as compared to Stochastic Gradient methods. K-FAC (Martens and Grosse, 2015) is a feasible implementation of NGD for DNNs which retains some desirable properties of NGD. Still, …
-
A Rock Borehole Packer System for Identifying Hydraulically Active Fractures Under Natural Gradient Flow
… of the borehole and its testing and not under natural flow conditions. The natural flow conditions are important because they govern contaminant transport in the whole flow system. Furthermore, conventional tracer tests are used to identify flow in fractures, but these too are typically done …
-
Superior to one of Glass: Natural Gradient index Lenses via Patchy Particle Self-Assembly
… suggests that lens proteins may form a gel with gradient density throughout the cellular lens structure, with density mediated by a tightly controlled protein coordination number in each region of the organ. Patchy colloid theory may therefore explain both the graded refractive index lens and the …
-
Discrete-depth measurements of natural gradient groundwater flow in fractures using borehole tracer methods
This thesis aims to quantify groundwater flow under ambient conditions in depth-discrete fractures using measurements in bedrock boreholes. The first method applied, known as the point dilution method, is well established and is performed in this study to serve as a benchmark for values of …
-
Natural Gradient Tracer Tests to Investigate the Fate and Migration of Oil Sands Process-Affected Water in the Wood Creek Sand Channel
… monitored to determine if naphthenic acids (NAs) naturally attenuated in the WCSC and if trace metals were mobilized from the aquifer solids due to changes in redox conditions. Post-injection monitoring found groundwater velocities through the aquifer were slow (~3-10 cm/day) despite hydraulic …
-
The effects of representing the spatial variability of aquifer characteristics on numerical ground water flow and contaminant transport modeling
… is dependent on the modeling purpose. In natural gradient ground water flow simulations the representation of hydraulic conductivity and bottom elevation significantly affected model calibration. The representation of aquifer properties generally did not affect model calibration of natural …
-
Geometry of Feedback Control and Learning
… directly, viewing control synthesis by policy gradient based algorithms. Adopting such a point of view has been partially inspired by the success of learning algorithms, such as Reinforcement Learning (RL), where using principles of Dynamic Programming (DP), one can devise real-time model-free …
-
Advancing High-Fidelity Neural Methods: Toward Fast and Accurate Solvers in Scientific Computing and Computational Fluid Dynamics
… accuracy even on smooth problems. Gauss–Newton Natural Gradient Descent (GNNG) addresses this by discretising Gauss–Newton’s method in function space, yielding a natural gradient update that provably mimics function-space dynamics and achieves close to single-precision accuracy on the …
-
Riemannian geometry based blind signal separation using independent component analysis
… the LMS-based algorithms to what is known as the natural-gradient update. The aim of this work is to generalize differential geometry algorithms for different cost functions of blind signal separation and provide new faster converging RLS-based and Newton-based algorithms using the natural …
-
Communication-efficient Second-order Optimization Methods on Modern Parallel Systems
… second-order methods such as quasi-Newtons and natural gradient descent is improved on modern computing systems. The sub-sampled quasi-Newton method is accelerated on distributed memory systems and demonstrates speedups of up to 12x compared to current efficient implementations. We develop, for …
-
Measuring groundwater flow system variability in a fractured carbonate aquifer using fibre optic distributed temperature sensing
… identified the location and magnitude of natural gradient flow within the aquifer sequence, and to what extent flow is affected by hydraulic cross-connection and pumping. Results show quantifiable changes in head and flow rates at discrete fractures to both hydraulic conditions. Variation …
-
Weak Solutions to a Fractional Fokker-Planck Equation via Splitting and Wasserstein Gradient Flow
… the presence of a potential field), there is no natural gradient flow formulation for its fractional counterpart, we prove existence of weak solutions to this fractional Fokker-Planck equation by combining a splitting technique together with a Wasserstein gradient flow formulation. An explicit …
-
A Novel Technique for Depth Discrete Flow Characterization: Fibre Optic Distributed Temperature Sensing within Boreholes Sealed with Flexible Underground Liners
… for identifying active groundwater flow under natural gradient conditions using DTS heat pulse testing is presented through a comparison with high resolution geologic logging and hydraulic datasets.
-
Three-dimensional simulation of the Columbus tritium plume
The migration of tritium during a large-scale natural-gradient tracer experiment in Columbus, Mississippi, was simulated using a three-dimensional finite element model for water flow and transport. The results showed that modeling approaches that assume complete homogeneity of the aquifer or of …
-
Geometry of Optimization in Markov Decision Processes and Neural Network-Based PDE Solvers
… algebraic techniques, and convex relaxations. Gradient-based methods, including variants of natural gradient methods, have gained tremendous attention in the theoretical reinforcement learning community, where they are commonly referred to as (natural) policy gradient methods. In Chapter 4, we …
-
Environmental Drivers of the Coral Reef Accretion-Erosion Balance in Present and Future Ocean Conditions
… eutrophication, and overfishing, as well as natural environmental variability. Predicting reef response to environmental stress requires an understanding of both natural and anthropogenic environmental drivers of reef accretion and erosion, and how these drivers interact at different …
Page 1 of 4