Search Results (4)

Using deep learning model predictions requires not only understanding the model’s confidence but also its uncertainty, so we know when to trust the prediction or require support from a human. In this study, we used Monte Carlo dropout (MCDO) to characterize the uncertainty of deep learning image classification algorithms, including feature fusion models, on simulated synthetic aperture radar (SAR) images of persistent ship wakes. Comparing to a baseline, we used the distribution of predictions from dropout with simple mean value ensembling and the Kolmogorov—Smirnov (KS) test to classify in-domain and out-of-domain (OOD) test samples, created by rotating images to angles not present in the training data. Our objective was to improve the classification robustness and identify OOD images during the test time. The mean value ensembling did not improve the performance over the baseline, in that there was a –1.05% difference in the Matthews correlation coefficient (MCC) from the baseline model averaged across all SAR bands. The KS test, by contrast, saw an improvement of +12.5% difference in MCC and was able to identify the majority of OOD samples. Leveraging the full distribution of predictions improved the classification robustness and allowed labeling test images as OOD. The feature fusion models, however, did not improve the performance over the single SAR-band models, demonstrating that it is best to rely on the highest quality data source available (in our case, C-band). Full article

Obesity remains one of the largest health problems in the world, arising from the excess storage of triglycerides (TAGs). However, the full complement of genes that are important for regulating TAG storage is not known. The Glut1 gene encodes a Drosophila glucose transporter that has been identified as a potential obesity gene through genetic screening. Yet, the tissue-specific metabolic functions of Glut1 are not fully understood. Here, we characterized the role of Glut1 in the fly brain by decreasing neuronal Glut1 levels with RNAi and measuring glycogen and TAGs. Glut1RNAi flies had decreased TAG and glycogen levels, suggesting a nonautonomous role of Glut1 in the fly brain to regulate nutrient storage. A group of hormones that regulate metabolism and are expressed in the fly brain are Drosophila insulin-like peptides (Ilps) 2, 3, and 5. Interestingly, we observed blunted Ilp3 and Ilp5 expression in neuronal Glut1RNAi flies, suggesting Glut1 functions in insulin-producing neurons (IPCs) to regulate whole-organism TAG and glycogen storage. Consistent with this hypothesis, we also saw fewer TAGs and glycogens and decreased expression of Ilp3 and Ilp5 in flies with IPC-specific Glut1RNAi. Together, these data suggest Glut1 functions as a nutrient sensor in IPCs, controlling TAG and glycogen storage and regulating systemic energy homeostasis. Full article

Training deep learning models requires having the right data for the problem and understanding both your data and the models’ performance on that data. Training deep learning models is difficult when data are limited, so in this paper, we seek to answer the following question: how can we train a deep learning model to increase its performance on a targeted area with limited data? We do this by applying rotation data augmentations to a simulated synthetic aperture radar (SAR) image dataset. We use the Uniform Manifold Approximation and Projection (UMAP) dimensionality reduction technique to understand the effects of augmentations on the data in latent space. Using this latent space representation, we can understand the data and choose specific training samples aimed at boosting model performance in targeted under-performing regions without the need to increase training set sizes. Results show that using latent space to choose training data significantly improves model performance in some cases; however, there are other cases where no improvements are made. We show that linking patterns in latent space is a possible predictor of model performance, but results require some experimentation and domain knowledge to determine the best options. Full article

Electrochemical deposition (ECD) is a common method used in the field of microelectronics to grow metallic coatings on an electrode. The deposition process occurs in an electrolyte bath where dissolved ions of the depositing material are suspended in an acid while an electric current is applied to the electrodes. The proposed computational model uses the finite volume method and the finite area method to predict copper growth on the plating surface without the use of a level set method or deforming mesh because the amount of copper layer growth is not expected to impact the fluid motion. The finite area method enables the solver to track the growth of the copper layer and uses the current density as a forcing function for an electric potential field on the plating surface. The current density at the electrolyte-plating surface interface is converged within each PISO (Pressure Implicit with Splitting Operator) loop iteration and incorporates the variance of the electrical resistance that occurs via the growth of the copper layer. This paper demonstrates the application of the finite area method for an ECD problem and additionally incorporates coupling between fluid mechanics, ionic diffusion, and electrochemistry. Full article