Spatially Explicit Hyperparameter Optimization for Neural Networks

1 online resource (154 pages) : PDF

2020

University of North Carolina at Charlotte

Neural networks as a commonly used machine learning algorithms, such as artificial neural networks (ANNs) and convolutional neural networks (CNNs), have been extensively used in GIScience domain to explore the nonlinear geographic phenomena. However, there are a few studies that investigate the parameter settings of neural networks in GIScience. Moreover, the model performance of neural networks often depends on the parameter setting for a given dataset. Meanwhile, adjusting the parameter configuration of neural networks will increase the overall running time. Therefore, an automated approach is necessary for addressing these limitations in current studies. In this dissertation, I proposed an automated spatially explicit hyperparameter optimization approach to identify optimal or near-optimal parameter settings for neural networks and accelerate the search process through both model and computing levels. I used two spatial prediction models in this dissertation to examine the utilities of spatially explicit hyperparameter optimization. The results demonstrate that the approach proposed in this dissertation improves the computing performance at model and computing levels and addresses the challenge of finding optimal parameter settings for neural networks in the GIScience field.

doctoral dissertations

Geodesy

Ph.D.

GIScience
Hyperparameter Optimization
Neural Networks
Parallel Computing

Geography & Urban Regional Analysis

Tang, Wenwu

Delmelle, Elizabeth
Lee, Minwoo
Ogundiran, Akin

Thesis (Ph.D.)--University of North Carolina at Charlotte, 2020.

This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s). For additional information, see http://rightsstatements.org/page/InC/1.0/.

Copyright is held by the author unless otherwise indicated.

Zheng_uncc_0694D_12511

http://hdl.handle.net/20.500.13093/etd:2587