This dissertation explores Random Neural Networks (RNNs) in several aspects and their applications. First, Novel RNNs have been proposed for dimensionality reduction and visualization. Based on Extreme Learning Machines (ELMs) and Self-Organizing Maps (SOMs) a new method is created to identify the important variables and visualize the data. This technique reduces the curse of dimensionality and improves furthermore the interpretability of the visualization and is tested on real nursing survey datasets. ELM-SOM+ is an autoencoder created to preserves the intrinsic quality of SOM and also brings continuity to the projection using two ELMs. This new methodology shows considerable improvement over SOM on real datasets. Second, as a Supervised Learning method, ELMs has been applied to the hierarchical multiscale method to bridge the the molecular dynamics to continua. The method is tested on simulation data and proven to be efficient for passing the information from one scale to another. Lastly, the regularization of ELMs has been studied and a new regularization algorithm for ELMs is created using a modified Lanczos Algorithm. The Lanczos ELM on average divide computational time by 20 and reduce the Normalized MSE by 14% comparing with regular ELMs.
Dissertation
Random neural networks for dimensionality reduction and regularized supervised learning
University of Iowa
Doctor of Philosophy (PhD), University of Iowa
Summer 2019
DOI: 10.17077/etd.d47y-9s7b
Abstract
Details
- Title: Subtitle
- Random neural networks for dimensionality reduction and regularized supervised learning
- Creators
- Renjie Hu - University of Iowa
- Contributors
- Amaury Lendasse (Advisor)Amany Farag (Committee Member)Yong Chen (Committee Member)Daniel McGehee (Committee Member)Edward Ratner (Committee Member)
- Resource Type
- Dissertation
- Degree Awarded
- Doctor of Philosophy (PhD), University of Iowa
- Degree in
- Industrial Engineering
- Date degree season
- Summer 2019
- Publisher
- University of Iowa
- DOI
- 10.17077/etd.d47y-9s7b
- Number of pages
- xii, 148 pages
- Copyright
- Copyright © 2019 Renjie Hu
- Language
- English
- Description illustrations
- illustrations (some color)
- Description bibliographic
- Includes bibliographical references (pages 134-148).
- Public Abstract (ETD)
This dissertation explores a few meaningful questions in Machine learning. The first question is: When we have a multi-feature dataset, which are the critical features, that preserve more information of the data than other features. Knowing these critical features will greatly improve the ability to interpret the data. A new method is created to identify the critical features and visualize them, by using Extreme Learning Machines (ELMs) and Self-Organizing Maps (SOMs). This method is tested on a real survey dataset of nurses in Chapter 2.
The second question is: When we have a multi-dimensional dataset, is it possible to reduce the dimensionality of the data but in the meantime preserve as much information as possible? In Chapter 4, A novel autoencoder is developed that conducts dimensionality reduction by preserving the data topology like SOMs but more importantly brings continuity to the projections using two ELMs.
The third question is: When we are training an ELM, how to select the complexity of the network? Usually, different datasets require different complexity of the ELMs. There is no way to know what complexity to choose before the training and validating process. In Chapter 5, a regularized ELM is proposed that could largely speed up the process of validating and allows ELM to have a large amount of hidden neurons without overfitting the problem.
- Academic Unit
- Industrial and Systems Engineering
- Record Identifier
- 9983776609402771
Metrics
399 File views/ downloads
139 Record Views