Search for: All records

The recent trend in regularization methods for inverse problems is to replace handcrafted sparsifying operators with datadriven approaches. Although using such machine learning techniques often improves image reconstruction methods, the results can depend significantly on the learning methodology. This paper compares two supervised learning methods. First, the paper considers a transform learning approach and, to learn the transform, introduces a variant on the Procrustes method for wide matrices with orthogonal rows. Second, we consider a bilevel convolutional filter learning approach. Numerical experiments show the learned transform performs worse for denoising than both the handcrafted finite difference transform and the learned filters, which perform similarly. Our results motivate the use of bilevel learning. 

This full “research” paper presents an overview of results of a systematic literature review of students' affective responses to active learning in undergraduate STEM courses. We considered 2,364 abstracts of conference papers and journal articles published since 1990, and 412 studies met our inclusion criteria. The studies span the STEM disciplines and report various types of active learning. Their research designs include primarily quantitative methods (especially instructor-designed surveys and course evaluations), and they find that students’ affective responses are overwhelmingly positive. Few studies excelled on our quality score metric, and there few statistically significant differences by discipline (but biology studies and chemistry studies scored significantly higher in quality than electrical engineering studies). We include several possible directions for future work. 

This work in progress paper presents an example of conducting a systematic literature review (SLR) to understand students’ affective response to active learning practices, and it focuses on the development and testing of a coding form for analyzing the literature. Specifically, the full paper seeks to answer: (1) what affective responses do instructors measure, (2) what evidence is used to study those responses, and (3) how are course features connected with student response. We conducted database searches with carefully-defined search queries which resulted in 2,365 abstracts from 1990 to 2015. Each abstract was screened by two researchers based on meeting inclusion criteria, with an adjudication round in the case of disagreement. We used RefWorks, an online citation management program, to track abstracts during this process. We identified over 480 abstracts which satisfied our criteria. Following abstract screening, we developed and tested a manuscript coding guide to capture the salient characteristics of each paper. We created an initial coding form by determining what paper topics would address our research questions and reviewing the literature to determine the most frequent response categories. We then piloted and tested the reliability of the form over three rounds of independent pair-coding, with each round resulting in clarifications to the form and mutual agreement on terms’ meanings. This process of developing a manuscript coding guide demonstrates how to use free online tools, such as Google Forms and Google Sheets, to inexpensively manage a large SLR team with significant turnover. Currently, we are in the process of applying the coding guide to the full texts. When complete, the resulting data will be synthesized by creating and testing relationships between variables, using each primary source as a case study to support or refute the hypothesized relationship.