Abstract
Optimization and machine learning are two of the most important developments in modern computational science. Over the years, research integrating these areas has proven to advance technology and find key insights for many applications. The interplay between these two areas has enjoyed prominence due to their ability to model real-life applications and designing algorithms that process large volumes of data at fast speeds.
One type of data that has gained attention in the past years are images. With an improvement in storage and processing capabilities, researchers are studying techniques to extract important features from two-dimensional information. Hence, this dissertation focuses on developing new algorithms that incorporate concepts from optimization and machine learning fields. Particularly, this dissertation presents novel techniques that cater effective solutions for two applications: Higher Order Aberration correction and power distribution management. In the first application, mathematical programming and image processing are utilized to improve vision for people with visual deformities. In the second application, machine learning and optimization are utilized to monitor and maintain vegetation surrounding the power grid.
This dissertation extensively researches the interaction of image processing, deep learning, and mathematical programming for developing cost-effective solutions with higher accuracy and lower computational complexity. Furthermore, the proposed methodologies illustrate the importance of multidisciplinary research and in-depth understanding of the literature across fields. These methodologies bridge the gap and act as an interface between electrical and industrial engineering and provide an outlet for developing creative and innovative solutions.