Doctor of Philosophy (PhD)
Recent advances in Artificial Intelligence and deep learning have provided researchers in various fields insights into the analysis of multiple datasets. These applications include image analysis, text analysis, and many more. However, the effectiveness of deep learning in some areas, such as biomedical imaging and genomic research, has been overshadowed by the variance in the types and complexity of data. This is in addition to the expensive labeling process and the limited size of datasets in these fields. These challenges require advanced deep learning models capable of learning from a small dataset and also from a small number of labeled data in unsupervised and semi-supervised fashion. As a result, the developed models for bioinformatics and biomedical data must be not only capable of taking advantage of unique software and high-performance computing environment but also capable of learning from heterogeneous datasets.
In developing effective deep learning algorithms capable of learning from heterogeneous datasets paired with the high-performance computing environment, we are able to not only analyze large and complex datasets but also make training and inference efficient.
In this dissertation, I study and develop multiple deep learning techniques for bioinformatics and biomedical applications. First, I illustrate a supervised deep learning model for lesion detection and breast cancer diagnosis from mammogram images. In the next chapter, I will discuss the limitation of previous work and improve it by using a generative adversarial network which supports transfer learning. Then, I illustrate how a generative model in a semi-supervised setting can be successfully applied to genomic data. Furthermore, I study the development of Bayesian deep learning and few-shot learning models to alleviate dependency on a large dataset and explore different methods to calculate uncertainty in breast cancer prediction from microscopic breast histopathology images.
Shams, Shayan, "A Study on Large-scale Deep Learning in Bioinformatics and Biomedical Applications" (2019). LSU Doctoral Dissertations. 4965.
Available for download on Saturday, June 11, 2022