Search for: All records

In this qualitative study, we ask: (1) How do Burundian girls and women describe their intersecting identities and (2) How do Burundian girls and women make decisions around STEM education and future careers? To answer these questions, we analyzed interviews conducted with eight Burundian families involved in a university-community organization partnership. 

The challenges faced by first-generation students, particularly within refugee communities, can be formidable as they aspire to attend an American university and pursue a professional career. These challenges include uncertainties in navigating the path from high school to college, limited awareness of various STEM career fields, and a lack of acquaintances who have successfully navigated similar paths. Complexities such as high school graduation and university admission requirements, coupled with few higher education connections, contribute to the frustrations experienced by parents and students. To address these issues, we present the results of a project aimed at promoting STEM aspirations, and enhancing the understanding of college navigation among refugee families residing in the United States. The project focused on parents and their children in grades 7-12 and was a collaboration between a large public university and leaders of several ethnic community-based organizations (ECBOs) representing local Burundian, Congolese, and Syrian communities. Results indicate the project positively affected students and parents’ STEM capital and college social capital, as well as students’ expectations regarding how fulfilling a STEM career might be. 

This research paper presents preliminary results of the Educational Ecosystem Health Survey (EEHS), a survey instrument designed by the Eco-STEM team at California State University, Los Angeles, a regionally serving, very high Hispanic-enrolling Minority Serving Institution (MSI). The purpose of the instrument is to quantitatively measure the health of the STEM educational ecosystem from the perspectives of the actors within it. The Eco-STEM team is implementing an ongoing NSF-funded research project aiming to change the paradigm of teaching and learning in STEM and its aligned mental models from factory-like to ecosystem- like. We hypothesize that this model of education will better support students and their individual needs. The pilot results of administering the EEHS to students within the College of Engineering, Computer Science, and Technology and the College of Natural and Social Sciences provide a baseline from which the Eco-STEM team will analyze diversion – and, hopefully, improvement – over the coming years of the project. The pilot survey was administered to undergraduate and graduate students at California State University, Los Angeles, of which the majority have ethnically- and socioeconomically- minoritized backgrounds. The EEHS is comprised of validated survey instruments that query students’ perceptions of various aspects of systemic educational health. These instruments measure the constructs of Classroom Comfort, Faculty Understanding, Belongingness, Thriving, Mindfulness, and Motivation. T-tests and ANOVA models are employed to analyze variations in responses among students based on a host of demographic identifiers. Pilot results from the first administration of the survey include, for example, statistically significant lower reported levels of thriving and mindfulness for students who identify as LGBTQIA+ than those who do not, as well as far lower levels of ecosystem health overall for students who do not have access to stable housing. Additional statistically significant results are identified on the bases of students’ gender, race/ethnicity, disability status, veteran status, undergraduate versus graduate student status, college of study, employment situation, and more detailed housing situation. The pilot results of the EEHS provide detailed insight into the experiences and needs of students in STEM programs at MSIs and regionally serving institutions. The results may also be useful within the contexts of a diverse range of institutions as they strive to serve students from historically marginalized backgrounds. 

Neural network based computer vision systems are typically built on a backbone, a pretrained or randomly initialized feature extractor. Several years ago, the default option was an ImageNet-trained convolutional neural network. However, the recent past has seen the emergence of countless backbones pretrained using various algorithms and datasets. While this abundance of choice has led to performance increases for a range of systems, it is difficult for practitioners to make informed decisions about which backbone to choose. Battle of the Backbones (BoB) makes this choice easier by benchmarking a diverse suite of pretrained models, including vision-language models, those trained via self-supervised learning, and the Stable Diffusion backbone, across a diverse set of computer vision tasks ranging from classification to object detection to OOD generalization and more. Furthermore, BoB sheds light on promising directions for the research community to advance computer vision by illuminating strengths and weakness of existing approaches through a comprehensive analysis conducted on more than 1500 training runs. While vision transformers (ViTs) and self-supervised learning (SSL) are increasingly popular, we find that convolutional neural networks pretrained in a supervised fashion on large training sets still perform best on most tasks among the models we consider. Moreover, in apples-to-apples comparisons on the same architectures and similarly sized pretraining datasets, we find that SSL backbones are highly competitive, indicating that future works should perform SSL pretraining. 

Neural network based computer vision systems are typically built on a backbone, a pretrained or randomly initialized feature extractor. Several years ago, the default option was an ImageNet-trained convolutional neural network. However, the recent past has seen the emergence of countless backbones pretrained using various algorithms and datasets. While this abundance of choice has led to performance increases for a range of systems, it is difficult for practitioners to make informed decisions about which backbone to choose. Battle of the Backbones (BoB) makes this choice easier by benchmarking a diverse suite of pretrained models, including vision-language models, those trained via self-supervised learning, and the Stable Diffusion backbone, across a diverse set of computer vision tasks ranging from classification to object detection to OOD generalization and more. Furthermore, BoB sheds light on promising directions for the research community to advance computer vision by illuminating strengths and weakness of existing approaches through a comprehensive analysis conducted on more than 1500 training runs. While vision transformers (ViTs) and self-supervised learning (SSL) are increasingly popular, we find that convolutional neural networks pretrained in a supervised fashion on large training sets still perform best on most tasks among the models we consider. Moreover, in apples-to-apples comparisons on the same architectures and similarly sized pretraining datasets, we find that SSL backbones are highly competitive, indicating that future works should perform SSL pretraining.