Deep-learning models have become pervasive tools in science and engineering. However, their energy requirements now increasingly limit their scalability1. Deep-learning accelerators2–9 aim to perform deep learning energy-efficiently, usually targeting the inference phase and often by exploiting physical substrates beyond conventional electronics. Approaches so far10–22 have been unable to apply the backpropagation algorithm to train unconventional novel hardware in situ. The advantages of backpropagation have made it the de facto training method for large-scale neural networks, so this deficiency constitutes a major impediment. Here we introduce a hybrid in situ–in silico algorithm, called physics-aware training, that applies backpropagation to train controllable physical systems. Just as deep learning realizes computations with deep neural networks made from layers of mathematical functions, our approach allows us to train deep physical neural networks made from layers of controllable physical systems, even when the physical layers lack any mathematical isomorphism to conventional artificial neural network layers. To demonstrate the universality of our approach, we train diverse physical neural networks based on optics, mechanics and electronics to experimentally perform audio and image classification tasks. Physics-aware training combines the scalability of backpropagation with the automatic mitigation of imperfections and noise achievable with in situ algorithms. Physical neural networks have the potential to perform machine learning faster and more energy-efficiently than conventional electronic processors and, more broadly, can endow physical systems with automatically designed physical functionalities, for example, for robotics23–26, materials27–29 and smart sensors30–32.
While there have been many calls to improve the quality of instructional physics labs, there exists little research on the effectiveness of lab instruction. This study provides a direct comparison between labs that have goals to reinforce physics content to those that emphasize experimentation skills. In this controlled study, all students attended the same lecture and discussion sections, had the same homework and exams, but attended labs that had one of two aims: teaching experimentation or reinforcing content. We compare students' engagement with experimentation during the lab as well as the impacts on students' exam performance and attitudes and beliefs about experimental physics. We find no measurable differences between lab conditions on students' exam performance. Nonetheless, we find measurable and significant improvements in students' engagement in expertlike experimentation practices and attitudes and beliefs about experimental physics for students in the experimentation labs. The benefits of the experimentation labs are stable across two subsequent semesters of implementation, as measured via standardized assessments. The results provide direct evidence of the extensive benefits of using labs to teach experimentation while directly demonstrating that shifting instructional goals and structure in labs can occur without cost to performance on course exams.
Many instructional physics labs are shifting to teach experimentation skills, rather than to demonstrate or confirm canonical physics phenomena. Our previous work found that many students engage in questionable research practices in attempts to confirm the canonical physics phenomena, even when confirmation is explicitly not the goal of the lab. This exploratory study aimed to answer three research questions: (RQ1) What are students' expectations about the purpose of labs when they enter introductory physics?, (RQ2) How do their prior experiences shape those expectations?, (RQ3) In what ways do those expectations relate to their engagement in questionable research practices? Through open-response surveys, we found that students overwhelmingly expressed confirmatory beliefs about the purpose of labs. Through interviews, we found that students' prior lab experiences were also overwhelmingly confirmatory, despite varying degrees of structure. We then used video of individual groups to explore the ways in which questionable research practices manifest through confirmatory expectations. We confirm previous work that students' confirmatory expectations can lead them to engage in questionable research practices, but find that these behaviors occur despite instructional messaging about an alternative purpose. Our analyses also suggest that engagement in questionable research practices is more frequent than the previous results indicated through analysis of submitted lab notes. These results further illuminate issues with traditional labs, but suggest that the confirmatory goals, perhaps more so than high structure, are problematic.
Framing affects how students interpret, approach, and accomplish tasks. Little is known, however, about how students frame tasks in physics labs. During the first lab of a sequence designed to teach students about modeling and critical thinking with data, students test a simple model of a pendulum that breaks down with improved measurements. Using in-lab video and follow-up interviews, we identified students' frequent use of a model-verifying frame that substantially interferes with the instructional goals. We present a case study analysis of two students who approach the lab with a model-verifying frame, engage in problematic behaviors including questionable research practices, but later shift their frames to accommodate goals aligned with instructional intention. As instructors transition their instructional labs to open-inquiry experiences, an activity that directly challenges the model-verifying frame may be productive for shifting students away from this problematic frame to one that supports their engagement in authentic experimentation.
Many institutions are changing the focus of their introductory physics labs from verifying physics content towards teaching students about the skills and nature of science. As instruction shifts, so too will the ways students approach and behave in the labs. In this study, we evaluated students' lab notes from an early activity in an experimentation-focused lab course. We found that about 30% of student groups (out of 107 groups at three institutions) recorded questionable research practices in their lab notes, such as subjective interpretations of results or manipulating equipment and data. The large majority of these practices were associated with confirmatory goals, which we suspect stem from students' prior exposure to verification labs. We propose ways for experimentation-focused labs to better engage students in the responsible conduct of research and authentic scientific practice.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.