Date of Award

8-1-2022

Language

English

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

College/School/Department

Department of Computer Science

Content Description

1 online resource (xvii, 106 pages) : illustrations (some color)

Dissertation/Thesis Chair

Pradeep K. Atrey

Committee Members

K. Narayanaswamy, Paliath Narendran, Shaghayegh Sahebi

Keywords

Active Learning, Deep Learning, Evidential Uncertainty, Genetic Algorithm, Lake Water Quality, Mushrooms Toxicity, Deep learning (Machine learning), Machine learning, Water quality, Crop science

Subject Categories

Artificial Intelligence and Robotics | Computer Sciences

Abstract

Despite significant advancements in the field of machine learning, there are two issues that still require further exploration. First, how to learn from a small dataset; and second, how to select appropriate features from the data. Although there exist many techniques to address these issues, choosing a combination of the techniques from these two groups is challenging, and worth investigating. To address these concerns, this thesis presents a learning framework that is based on a deep learning model utilizing active learning (with evidential uncertainty as a basis for acquisition function) for the first issue and a genetic algorithm for the second. The framework is named deep active genetic with evidential uncertainty, abbreviated as DAG-EU. The proposed DAG-EU framework has the following characteristics: it can handle problems with a smaller dataset and a large number of features; does not require extra knowledge about the problem domain; can remove redundant features which add noise to the model and make model interpretation difficult and problematic; and it can choose its observations to learn from, avoid noisy ones and keep higher model performance. Furthermore, the performance of our model is improved by including subjective logic to measure model uncertainty through evidential uncertainty. For this, the thesis presents a new acquisition function that is used to select data points, which the model is most uncertain about their labels.

Download

Share

COinS