Title: Pull-back Geometry of Persistent Homology Encodings
Persistent homology (PH) is a method for generating topology-inspired representations of data. Empirical studies that investigate the properties of PH, such as its sensitivity to perturbations or ability to detect a feature of interest, commonly rely on training and testing an additional model on the basis of the PH representation. To gain more intrinsic insights about PH, independently of the choice of such a model, we propose a novel methodology based on the pull-back geometry that a PH encoding induces on the data manifold. The spectrum and eigenvectors of the induced metric help to identify the most and least significant information captured by PH. Furthermore, the pull-back norm of tangent vectors provides insights about the sensitivity of PH to a given perturbation, or its potential to detect a given feature of interest, and in turn its ability to solve a given classification or regression problem. Experimentally, the insights gained through our methodology align well with the existing knowledge about PH. Moreover, we show that the pull-back norm correlates with the performance on downstream tasks, and can therefore guide the choice of a suitable PH encoding. more »« less
Buscombe, Daniel; Carini, Roxanne J.
(, Remote Sensing)
null
(Ed.)
We apply deep convolutional neural networks (CNNs) to estimate wave breaking type (e.g., non-breaking, spilling, plunging) from close-range monochrome infrared imagery of the surf zone. Image features are extracted using six popular CNN architectures developed for generic image feature extraction. Logistic regression on these features is then used to classify breaker type. The six CNN-based models are compared without and with augmentation, a process that creates larger training datasets using random image transformations. The simplest model performs optimally, achieving average classification accuracies of 89% and 93%, without and with image augmentation respectively. Without augmentation, average classification accuracies vary substantially with CNN model. With augmentation, sensitivity to model choice is minimized. A class activation analysis reveals the relative importance of image features to a given classification. During its passage, the front face and crest of a spilling breaker are more important than the back face. For a plunging breaker, the crest and back face of the wave are most important, which suggests that CNN-based models utilize the distinctive ‘streak’ temperature patterns observed on the back face of plunging breakers for classification.
We present a MEMS microphone that converts the mechanical motion of a diaphragm, generated by acoustic waves, to an electrical output voltage by capacitive fingers. The sensitivity of a microphone is one of the most important properties of its design. The sensitivity is proportional to the applied bias voltage. However, it is limited by the pull-in voltage, which causes the parallel plates to collapse and prevents the device from functioning properly. The presented MEMS microphone is biased by repulsive force instead of attractive force to avoid pull-in instability. A unit module of the repulsive force sensor consists of a grounded moving finger directly above a grounded fixed finger placed between two horizontally seperated voltage fixed fingers. The moving finger experiences an asymmetric electrostatic field that generates repulsive force that pushes it away from the substrate. Because of the repulsive nature of the force, the applied voltage can be increased for better sensitivity without the risk of pull-in failure. To date, the repulsive force has been used to engage a MEMS actuator such as a micro-mirror, but we now apply it for a capacitive sensor. Using the repulsive force can revolutionize capacitive sensors in many applications because they will achieve better sensitivity. Our simulations show that the repulsive force allows us to improve the sensitivity by increasing the bias voltage. The applied voltage and the back volume of a standard microphone have stiffening effects that significantly reduce its sensitivity. We find that proper design of the back volume and capacitive fingers yield promising results without pull-in instability.
Ozdogan, Mehmet; Towfighian, Shahrzad; Miles, Ronald N.
(, 2019 IEEE SENSORS)
Fabrication and acoustic performance of a microelectromechanical systems (MEMS) microphone are presented. The microphone utilizes an unusual electrostatic sensing scheme that causes the sensing electrode to move away, or levitate from the biasing electrode as the bias voltage is applied. This approach differs from existing electrostatic sensors and completely avoids the usual collapse, or pull-in instability. In this study, our goal is to fabricate a MEMS microphone whose sensitivity could be improved simply by increasing the bias voltage, without suffering from pull-in instability. The microphone is tested in our anechoic chamber and a read-out circuit is used to obtain electrical signals in response to sound pressure at various bias voltages. Experimental results show that the sensitivity increases approximately linearly with bias voltage for bias voltages from 40 volts to 100 volts. The ability to design electrostatic sensors without concerns about pull-in failure can enable a wide-range of promising sensor designs.
The ability to study and visualize metabolites on a cellular and sub-cellular level is important for gaining insights into biological pathways and metabolism of multicellular organisms. Matrix-assisted laser desorption/ionization (MALDI) mass spectrometry imaging (MSI) is a powerful analytical tool for metabolomics experiments due to its high sensitivity and small sampling size. The spatial resolution in MALDI-MSI is mainly limited by the number of molecules available in a small sampling size. When the sampling size is low enough to achieve cellular or subcellular spatial resolution, signal intensity is sacrificed making poorly ionized metabolites difficult to detect. To overcome this limitation, on-tissue chemical derivatization reactions have been used to enhance the desorption/ionization efficiency of selected classes of compounds by adding a functional group with a permanent positive charge or one that can be easily ionized. By utilizing several chemical derivatizations in parallel, metabolite coverage can be drastically improved. This chapter outlines methodology for sample preparation and data analysis for on-tissue chemical derivatization using various derivatization reagents.
Vaziri, S.L.; Paretti, M.C.; Grohs, J.R.; Baum, L.M.; Lester, M.M.; Newbill, P.L.
(, Zone 1 Conference of the American Society for Engineering Education)
Broadening participation in engineering is critical given the gap between the nation’s need for engineering graduates and its production of them. Efforts to spark interest in engineering among PreK-12 students have increased substantially in recent years as a result. However, past research has demonstrated that interest is not always sufficient to help students pursue engineering majors, particularly for rural students. In many rural communities, influential adults (family, friends, teachers) are often the primary influence on career choice, while factors such as community values, lack of social and cultural capital, limited course availability, and inadequate financial resources act as potential barriers. To account for these contextual factors, this project shifts the focus from individual students to the communities to understand how key stakeholders and organizations support engineering as a major choice and addresses the following questions: RQ1. What do current undergraduate engineering students who graduated from rural high schools describe as influences on their choice to attend college and pursue engineering as a post-secondary major? RQ2. How does the college choice process differ for rural students who enrolled in a 4-year university immediately after graduating from high school and those who transferred from a 2-year institution? RQ3. How do community members describe the resources that serve as key supports as well as the barriers that hinder support in their community? RQ4. What strategies do community members perceive their community should implement to enhance their ability to support engineering as a potential career choice? RQ5. How are these supports transferable or adaptable by other schools? What community-level factors support or inhibit transfer and adaptation? To answer the research questions, we employed a three-phase qualitative study. Phase 1 focused on understanding the experiences and perceptions of current [University Name] students from higher-producing rural schools. Analysis of focus group and interview data with 52 students highlighted the importance of interest and support from influential adults in students’ decision to major in engineering. One key finding from this phase was the importance of community college for many of our participants. Transfer students who attended community college before enrolling at [University Name] discussed the financial influences on their decision and the benefits of higher education much more frequently than their peers. In Phase 2, we used the findings from Phase 1 to conduct interviews within the participants’ home communities. This phase helped triangulate students’ perceptions with the perceptions and practices of others, and, equally importantly, allowed us to understand the goals, attitudes, and experiences of school personnel and local community members as they work with students. Participants from the students’ home communities indicated that there were few opportunities for students to learn more about engineering careers and provided suggestions for how colleges and universities could be more involved with students from their community. Phase 3, scheduled for Spring 2020, will bring the findings from Phases 1 and 2 back to rural communities via two participatory design workshops. These workshops, designed to share our findings and foster collaborative dialogue among the participants, will enable us to explore factors that support or hinder transfer of findings and to identify policies and strategies that would enhance each community’s ability to support engineering as a potential career choice.
Liang, Shuang, Turkes, Renata, Li, Jiayi, Otter, Nina, and Montufar, Guido.
"Pull-back Geometry of Persistent Homology Encodings". Transactions on Machine Learning Research (). Country unknown/Code not available: TMLR. https://par.nsf.gov/biblio/10515009.
@article{osti_10515009,
place = {Country unknown/Code not available},
title = {Pull-back Geometry of Persistent Homology Encodings},
url = {https://par.nsf.gov/biblio/10515009},
abstractNote = {Persistent homology (PH) is a method for generating topology-inspired representations of data. Empirical studies that investigate the properties of PH, such as its sensitivity to perturbations or ability to detect a feature of interest, commonly rely on training and testing an additional model on the basis of the PH representation. To gain more intrinsic insights about PH, independently of the choice of such a model, we propose a novel methodology based on the pull-back geometry that a PH encoding induces on the data manifold. The spectrum and eigenvectors of the induced metric help to identify the most and least significant information captured by PH. Furthermore, the pull-back norm of tangent vectors provides insights about the sensitivity of PH to a given perturbation, or its potential to detect a given feature of interest, and in turn its ability to solve a given classification or regression problem. Experimentally, the insights gained through our methodology align well with the existing knowledge about PH. Moreover, we show that the pull-back norm correlates with the performance on downstream tasks, and can therefore guide the choice of a suitable PH encoding.},
journal = {Transactions on Machine Learning Research},
publisher = {TMLR},
author = {Liang, Shuang and Turkes, Renata and Li, Jiayi and Otter, Nina and Montufar, Guido},
}
Warning: Leaving National Science Foundation Website
You are now leaving the National Science Foundation website to go to a non-government website.
Website:
NSF takes no responsibility for and exercises no control over the views expressed or the accuracy of
the information contained on this site. Also be aware that NSF's privacy policy does not apply to this site.
An official website of the United States government
