We report an experimental measurement of the acoustic signal emitted from an individual suspended carbon nanotube (CNT) approximate 2 μm in length, 1 nm in diameter, and 10 À21 kg in mass. This system represents the smallest thermoacoustic system studied to date. By applying an AC voltage of 1.4 V at 8 kHz to the suspended CNT, we are able to detect the acoustic signal using a commercial microphone. The acoustic power detected is found to span a range from 0.1 to 2.4 attoWatts or 0.2 to 1 μPa of sound pressure. This corresponds to thermoacoustic efficiencies ranging from 0.007 to 0.6 Pa/W for the seven devices that were measured in this study. Here, the small lateral dimensions of these devices cause large heat losses due to thermal conduction, which result in the relatively small observed thermoacoustic efficiencies.
A surprising phenomenon in modern machine learning is the ability of a highly overparameterized model to generalize well (small error on the test data) even when it is trained to memorize the training data (zero error on the training data). This has led to an arms race towards increasingly overparameterized models (c.f., deep learning). In this paper, we study an underexplored hidden cost of overparameterization: the fact that overparameterized models are more vulnerable to privacy attacks, in particular the membership inference attack that predicts the (potentially sensitive) examples used to train a model. We significantly extend the relatively few empirical results on this problem by theoretically proving for an overparameterized linear regression model with Gaussian data that the membership inference vulnerability increases with the number of parameters. Moreover, a range of empirical studies indicates that more complex, nonlinear models exhibit the same behavior. Finally, we study different methods for mitigating such attacks in the overparameterized regime, such as noise addition and regularization, and conclude that simply reducing the parameters of an overparameterized model is an effective strategy to protect it from membership inference without greatly decreasing its generalization error.
Visual representations are prevalent in STEM instruction. To benefit from visuals, students need representational competencies that enable them to see meaningful information. Most research has focused on explicit conceptual representational competencies, but implicit perceptual competencies might also allow students to efficiently see meaningful information in visuals. Most common methods to assess students’ representational competencies rely on verbal explanations or assume explicit attention. However, because perceptual competencies are implicit and not necessarily verbally accessible, these methods are ill‐equipped to assess them. We address these shortcomings with a method that draws on similarity learning, a machine learning technique that detects visual features that account for participants’ responses to triplet comparisons of visuals. In Experiment 1, 614 chemistry students judged the similarity of Lewis structures and in Experiment 2, 489 students judged the similarity of ball‐and‐stick models. Our results showed that our method can detect visual features that drive students’ perception and suggested that students’ conceptual knowledge about molecules informed perceptual competencies through top‐down processes. Furthermore, Experiment 2 tested whether we can improve the efficiency of the method with active sampling. Results showed that random sampling yielded higher accuracy than active sampling for small sample sizes. Together, the experiments provide the first method to assess students’ perceptual competencies implicitly, without requiring verbalization or assuming explicit visual attention. These findings have implications for the design of instructional interventions that help students acquire perceptual representational competencies.
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.