Self-attention transformers have demonstrated accuracy for image classification with smaller data sets. However, a limitation is that tests to-date are based upon single class image detection with known representation of image populations. For instances where the input image classes may be greater than one and test sets that lack full information on representation of image populations, accuracy calculations must adapt. The Receiver Operating Characteristic (ROC) accuracy thresh-old can address the instances of multi-class input images. However, this approach is unsuitable in instances where image population representation is unknown. We consider calculating accuracy using the knee method to determine threshold values on an ad-hoc basis. Results of ROC curve and knee thresholds for a multi-class data set, created from CIFAR-10 images, are discussed for multi-class image detection.
more »
« less
This content will become publicly available on May 15, 2025
"Len or index or count, anything but v1": Predicting Variable Names in Decompilation Output with Transfer Learning
More Like this
-
-
more » « less
Abstract The Fixed Anvil Temperature (FAT) hypothesis proposes that upper tropospheric cloud fraction peaks at a special isotherm that is independent of surface temperature. It has been argued that a FAT should result from simple ingredients: Clausius‐Clapeyron, longwave emission from water vapor, and tropospheric energy and mass balance. Here the first cloud‐resolving simulations of radiative‐convective equilibrium designed to contain only these basic ingredients are presented. This setup does not produce a FAT: the anvil temperature varies by about 40% of the surface temperature range. However, the tropopause temperature varies by only 4% of the surface temperature range, which supports the existence of a Fixed Tropopause Temperature (FiTT). In full‐complexity radiative‐convective equilibrium simulations, the spread in anvil temperature is smaller by about a factor of 2, but the tropopause temperature remains more invariant than the anvil temperature by an order of magnitude. In other words, our simulations have a FiTT, not a FAT.
-
Abstract In this study, we investigated an unusual natural Mn oxide hollandite-group mineral from the Kohare Mine, Iwate Prefecture, Japan, that has predominantly water molecules in the tunnels, with K, Na, Ca, and Ba. The specimens are labeled as type manjiroite, but our analyses show that Na is not the dominant tunnel species, nor is it even the primary tunnel cation, suggesting either an error in the original analyses or significant compositional variation within samples from the type locality. Chemical analyses, X-ray photoelectron spectroscopy, and thermal gravimetric analysis measurements combined with Rietveld refinement results using synchrotron X-ray powder diffraction data suggest the chemical formula: (K0.19Na0.17Ca0.03Ba0.01H2O1.60)(Mn5.024+Mn2.823+Al0.14Fe0.02)O13.47(OH)2.53. Our analyses indicate that water is the primary tunnel species, and although water has been reported as a component in natural hollandites, this is the first detailed study of the crystal structure and dehydration behavior of a natural hydrous hollandite with water as the predominant tunnel species. This work underscores the rarity of natural Na-rich hollandite phases and focuses new attention on the role of hydrous components of hollandite-like phases in determining their capacities to exchange or accommodate various cations, such as Li+, Na+, Ba2+, Pb2+, and K+ in natural systems.more » « less
