On Leveraging Representation Learning Techniques for Data Analytics in Biomedical Informatics
Genre
Thesis/DissertationDate
2019Author
Cao, Xi HangAdvisor
Obradovic, ZoranCommittee member
Vucetic, SlobodanSouvenir, Richard M.
Kaplan, Avi
Department
Computer and Information ScienceSubject
Computer ScienceArtificial Intelligence
Information Science
Data Analytics
Data Mining
Machine Learning
Representation Learning
Supervised Learning
Unsupervised Learning
Permanent link to this record
http://hdl.handle.net/20.500.12613/903
Metadata
Show full item recordDOI
http://dx.doi.org/10.34944/dspace/885Abstract
Representation Learning is ubiquitous in state-of-the-art machine learning workflow, including data exploration/visualization, data preprocessing, data model learning, and model interpretations. However, the majority of the newly proposed Representation Learning methods are more suitable for problems with a large amount of data. Applying these methods to problems with a limited amount of data may lead to unsatisfactory performance. Therefore, there is a need for developing Representation Learning methods which are tailored for problems with ``small data", such as, clinical and biomedical data analytics. In this dissertation, we describe our studies of tackling the challenging clinical and biomedical data analytics problem from four perspectives: data preprocessing, temporal data representation learning, output representation learning, and joint input-output representation learning. Data scaling is an important component in data preprocessing. The objective in data scaling is to scale/transform the raw features into reasonable ranges such that each feature of an instance will be equally exploited by the machine learning model. For example, in a credit flaw detection task, a machine learning model may utilize a person's credit score and annual income as features, but because the ranges of these two features are different, a machine learning model may consider one more heavily than another. In this dissertation, I thoroughly introduce the problem in data scaling and describe an approach for data scaling which can intrinsically handle the outlier problem and lead to better model prediction performance. Learning new representations for data in the unstandardized form is a common task in data analytics and data science applications. Usually, data come in a tubular form, namely, the data is represented by a table in which each row is a feature (row) vector of an instance. However, it is also common that the data are not in this form; for example, texts, images, and video/audio records. In this dissertation, I describe the challenge of analyzing imperfect multivariate time series data in healthcare and biomedical research and show that the proposed method can learn a powerful representation to encounter various imperfections and lead to an improvement of prediction performance. Learning output representations is a new aspect of Representation Learning, and its applications have shown promising results in complex tasks, including computer vision and recommendation systems. The main objective of an output representation algorithm is to explore the relationship among the target variables, such that a prediction model can efficiently exploit the similarities and potentially improve prediction performance. In this dissertation, I describe a learning framework which incorporates output representation learning to time-to-event estimation. Particularly, the approach learns the model parameters and time vectors simultaneously. Experimental results do not only show the effectiveness of this approach but also show the interpretability of this approach from the visualizations of the time vectors in 2-D space. Learning the input (feature) representation, output representation, and predictive modeling are closely related to each other. Therefore, it is a very natural extension of the state-of-the-art by considering them together in a joint framework. In this dissertation, I describe a large-margin ranking-based learning framework for time-to-event estimation with joint input embedding learning, output embedding learning, and model parameter learning. In the framework, I cast the functional learning problem to a kernel learning problem, and by adopting the theories in Multiple Kernel Learning, I propose an efficient optimization algorithm. Empirical results also show its effectiveness on several benchmark datasets.ADA compliance
For Americans with Disabilities Act (ADA) accommodation, including help with reading this content, please contact scholarshare@temple.eduCollections
Related items
Showing items related by title, author, creator and subject.
-
WRITING TO LEARN STRATEGIES ON STUDENT LEARNING IN BIOLOGY: AN EXPERIMENT WITH STUDENTS WITH SPECIFIC LEARNING DISABILITESDuCette, Joseph P.; Tincani, Matt; DuCette, Joseph P.; Tincani, Matt; Thurman, S. Kenneth; Sawaya, Bassel E. (Temple University. Libraries, 2020)Current research has shown that writing to learn strategies have a positive impact on overall student learning. However, previous studies mainly assessed student learning through instruments that have been self-designed by researchers and there is little research examining relations between writing to learn strategies and science related context achievement such as biology measured through standardized testing. In addition, very few studies on writing to learn strategies with students with disabilities measure the potential relations with biology achievement. The purpose of this quantitative study is to address the literature gap by implementing the use of writing to learn strategies to institute growth and achievement in biological science in students with disabilities. The objective of the study is to evaluate content area growth of students with disabilities in biological science through the use of writing to learn strategies embedded in classroom pedagogy. Using experimental research, forty-eight students with high incidence disabilities in grades 9 and 10 received writing to learn strategies in their biology class during the school day. The quantitative instrumentation involved in the study is a pre-test comprehensive biology baseline exam, a mid-treatment biology assessment, and a post-test comprehensive biology exam, an acceptability survey, and an attitude toward biology survey that was administered over a semester’s timeline. The exams simulated standardized biology exams that employ the New Jersey Science Standards (NJSS) taught within the curriculum. All exams were slightly different; however, they focused on the same standards in biological science and the same overall content. Results were that the experimental group showed greater growth than the control group in the mid-treatment assessment, posttest and the Attitude Toward Biology Survey. Empirical implications are that writing to learn strategies would have significant impact on biological science standardized testing for students with disabilities.
-
Use Of A Trauma Informed Care Framework To Create Bidirectional Learning Opportunities In A Critical Service Learning CurriculumJones, Nora L. (Temple University. Libraries, 2020)Addressing childhood trauma is increasingly being recognized as a priority in public health, healthcare and health policy sectors. As evidence mounts that the effects of trauma are both graded and dose responsive, stakeholders in healthcare are turning more attention to preventing and addressing experiences of trauma in childhood, commonly referred to as adverse childhood experiences (ACEs). Trauma Informed Care (TIC), is a promising clinical approach attuned to the specific needs of traumatized patients that is still in its infancy with regards to training and evaluation of practitioner skills. Although physicians of all specialties will encounter patients who have experienced trauma, few undergraduate medical education curriculums provide formal training in TIC. Additionally, the approach to TIC in clinical settings has largely been biomedical and individual, with a focus on screening and treatment. This model ignores the role that communities play in both propagating trauma and generating resiliency. Urban academic medical centers, often geographically located in highly traumatized communities, must take a specific interest in developing TIC research, theory and praxis that includes and empowers communities. Service learning, a form of experiential education that cultivates self-awareness in students while simultaneously meeting community objectives, is a pedagogy that aligns with a community driven TIC framework. This thesis outlines the implementation of a community driven, bi-directional TIC learning model designed to serve the needs of medical students and low income K-8th grade students living in the geographic catchment of an urban, academic medical center.
-
The Effects of a Combined Mathematics Content/Methods Course Involving Learning Cycles on Preservice Elementary Teachers' Beliefs about the Nature, Teaching and Learning of MathematicsLeonard, Jacqueline; DuCette, Joseph P.; Schmuckler, Joseph S., 1927-; Davis, James Earl, 1960-; Ryan, Steve (Temple University. Libraries, 2009)The purpose of this study was to determine whether a one-semester combined mathematics content/methods course based on Simon's (1994) Learning Cycle framework for the mathematical preparation of preservice elementary teachers was more effective at aligning preservice teachers' beliefs with the philosophy of the Standards documents (NCTM, 1989, 1991, 1995, 2000) and other reform literature than another content/methods course. While the six-cycled Learning Cycles framework has been used in separate content and methods courses (Simon, 1995), it has not been used in a combined content/methods setting. The preservice teachers enrolled in two sections of the course Teaching Math to Children: N-6 were involved in the study with one section investigating the topics measurement, geometry and fractions using the Learning Cycles framework while the other section used a different framework to investigate number concepts and place value; addition and subtraction strategies; multiplication and division strategies; geometry and spatial sense; ratio, proportion, percent, similarity and scale; and probability, counting techniques and statistics. Beliefs about the nature, teaching and learning of mathematics were measured both quantitatively and qualitatively. All participants' beliefs were assessed quantitatively at the beginning and the end of the course using a beliefs instrument with three subscales that measured beliefs about the nature, teaching and learning of mathematics. Five preservice teachers from the Learning Cycles course served as case studies and provided qualitative data through interviews and journal prompts. Analyses of Covariance were calculated for course of enrollment and each of the three beliefs subscales. For beliefs about the nature of mathematics, there was a significant difference between the adjusted post-test scores of the Learning Cycles group and the comparison group favoring the Learning Cycles group. Similar results were found regarding beliefs about the teaching of mathematics. While significant differences were found between the groups with regards to beliefs about the learning of mathematics, the scores on the corresponding subscale changed only slightly. Qualitative analyses confirmed the findings regarding the beliefs about the nature and teaching of mathematics. Differences were also found in the preservice teachers' beliefs about learning mathematics. The case study participants attributed certain characteristics of the course with their changing beliefs. These results indicate that the Learning Cycles course was somewhat successful at aligning preservice teachers' beliefs with those of the reform movement.