David D. Allred received Alumni Professorship Award

Maximum entropy is an approach for obtaining posterior probability distributions of modeling parameters. This approach, based on a cost function that quantifies the data-model mismatch, relies on an estimate of an appropriate temperature. Selection of this ”statistical temperature” is related to estimating the noise covariance. A method for selecting the ”statistical temperature” is derived from analogies with statistical mechanics, including the equipartition theorem. Using the equipartition-theorem estimate, the statistical temperature can be obtained for a single data sample instead of via the ensemble approach used previously. Examples of how the choice of temperature impacts the posterior distributions are shown using a toy model. The examples demonstrate the impact of the choice of the temperature on the resulting posterior probability distributions and the advantages of using the equipartition-theorem approach for selecting the temperature.

The information contained in spectrograms of ocean shipping noise is explored with a deep learning method for seabed classification. Due to the lack of measured and labeled ship noise data, the residual convolutional neural network ResNet-18 is trained with synthetic data samples generated by physics-based models. Previously, spectrograms from a single hydrophone channel, each scaled by their standard deviation, were utilized to infer an effective seabed class. This work, in contrast, utilizes multiple channels and a new scaling of the data more suited to deep learning. Models are trained with data from one, two, or four channels corresponding to hydrophones of different depths. The validation results show that for two-channel models, models with data from hydrophones at a lower average depth perform slightly better. Model performance is shown to depend neither on whether a specific channel is included nor the distance between channels in the model. These conclusions allow future multichannel studies to narrow down the number of models needed to obtain comprehensive results.

Complementary metal-oxide-semiconductor high-density microelectrode array (CMOS-HD-MEA) systems can record neurophysiological activity from cell cultures and ex vivo brain slices in unprecedented electrophysiological detail. CMOS-HD-MEAs were first optimized to record high-quality neuronal unit activity from cell cultures but have also been shown to produce quality data from acute retinal and cerebellar slices. Researchers have recently used CMOS-HD-MEAs to record local field potentials (LFPs) from acute, cortical rodent brain slices. One LFP of interest is seizure-like activity. While many users have produced brief, spontaneous epileptiform discharges using CMOS-HD-MEAs, few users reliably produce quality seizure-like activity. Many factors may contribute to this difficulty, including electrical noise, the inconsistent nature of producing seizure-like activity when using submerged recording chambers, and the limitation that 2D CMOS-MEA chips only record from the surface of the brain slice. The techniques detailed in this protocol should enable users to consistently induce and record high-quality seizure-like activity from acute brain slices with a CMOS-HD-MEA system. In addition, this protocol outlines the proper treatment of CMOS-HD-MEA chips, the management of solutions and brain slices during experimentation, and equipment maintenance.

Undergraduate students on track for medical school are often required to take general physics lab courses. Many of these students carry an attitude of obligation into these courses which can make it challenging for instructors to engage students in course material. We address the question: How does engaging with medically based models in introductory physics labs affect pre-med undergraduate perceptions of the modeling process and their perceptions of science? We redesigned an electricity and magnetism lab in an introductory physics lab course, where approximately 70% of the undergraduates reported plans to attend medical school. We situated the lab in the mechanics of MRI magnetic resonance and collected data on the participants’ experiences through surveys and lab submissions. As a part of the analysis, we modified a rubric to evaluate engagement in modeling and applied grounded coding theory to the survey responses to develop themes of the participants’ understanding of scientific modeling. The participants’ understanding and engagement in scientific modeling increased during the newly developed lab and remained high for subsequent labs. We recommend that instructors of undergraduate nonmajor labs consider the demographic of their student population and design lab experiences situated within their interests and focus on central science practices like modeling.

Single-shot two-dimensional (2D) phase retrieval (PR) can recover the phase shift distribution within an object from a single 2D x-ray phase contrast image (XPCI). Two competing XPCI imaging modalities often used for single-shot 2D PR to recover material properties critical for predictive performance capabilities are: speckle-based (SP-XPCI) and propagation-based (PB-XPCI) XPCI imaging. However, PR from SP-XPCI and PB-XPCI images are, respectively, limited to reconstructing accurately slowly and rapidly varying features due to noise and differences in their contrast mechanisms. Herein, we consider a combined speckle- and propagation-based XPCI (SPB-XPCI) image by introducing a mask to generate a reference pattern and imaging in the near-to-holographic regime to induce intensity modulations in the image. We develop a single-shot 2D PR method for SPB-XPCI images of pure phase objects without imposing restrictions such as object support constraints. It is compared against PR methods inspired by those developed for SP-XPCI and PB-XPCI on simulated and experimental images of a thin glass shell before and during shockwave compression. Reconstructed phase maps show improvements in quantitative scores of root-mean-square error and structural similarity index measure using our proposed method.

Group-theoretical and linear-algebraic methods and tools have recently been developed that aim to exhaustively identify the small-angle rotational rigid-unit modes (RUMs) of a given framework material. But in their current form, they fail to detect RUMs that require a compensating lattice strain which grows linearly with the amplitude of the rigid-unit rotations. Here, we present a systematic approach to including linear strain compensation within the linear-algebraic RUM-search method, so that any geometrically possible small-angle RUM can be detected.