Humans can prepare concise descriptions of pictures, focusing on what they find important. We demonstrate that automatic methods can do so too. We describe a system that can compute a score linking an image to a sentence. This score can be used to attach a descriptive sentence to a given image, or to obtain images that illustrate a given sentence. The score is obtained by comparing an estimate of meaning obtained from the image to one obtained from the sentence. Each estimate of meaning comes from a discriminative procedure that is learned using data. We evaluate on a novel dataset consisting of human-annotated images. While our underlying estimate of meaning is impoverished, it is sufficient to produce very good quantitative results, evaluated with a novel score that can account for synecdoche.
The redox flow battery is a promising energy storage technology for managing the inherent uncertainty of renewable energy sources. At present, however, they are too expensive and thus economically unattractive. Optimizing flow batteries is thus an active area of research, with the aim of reducing cost by maximizing performance. This work addresses microstructural electrode optimizations by providing a modeling framework based on pore-networks to study the multiphysics involved in a flow battery, with a specific focus on pore-scale structure and its impact on transport processes. The proposed pore network approach was extremely cheap in computation cost (compared to direct numerical simulation) and therefore was used for parametric sweeps to search for optimum electrode structures in a reasonable time. It was found that that increasing porosity generally helps performance by increasing the permeability and flow rate at a given pressure drop, despite reducing reactive surface area per unit volume. As a more nuanced structural study, it was found that aligning fibers in the direction of flow helps performance by increasing permeability but showed diminishing returns beyond slight alignment. The proposed model was demonstrated in the context of a hydrogen bromine flow battery but could be applied to any system of interest.) unless CC License in place (see abstract). ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 128.41.35.179 Downloaded on 2019-07-09 to IP A2122 Journal of The Electrochemical Society, 166 (10) A2121-A2130 (2019) ) unless CC License in place (see abstract). ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 128.41.35.179 Downloaded on 2019-07-09 to IP ) unless CC License in place (see abstract). ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 128.41.35.179 Downloaded on 2019-07-09 to IP
A numerical method for calculating the mass transfer coefficient in fibrous media is presented. First, pressure driven flow was modelled using the Lattice Boltzmann Method. The advection-diffusion equation was solved for convective-reacting porous media flow, and the method is contrasted with experimental methods such as the limiting current diffusion technique, for its ability to determine and simulate mass transfer systems that are operating at low Reynolds number flows. A series of simulations were performed on three materials; specifically, commercially available carbon felts, electrospun carbon fibers and electrospun carbon fibers with anisotropy introduced to the microstructure. Simulations were performed in each principal direction (x,y,z) for each material in order to determine the effects of anisotropy on the mass transfer coefficient. In addition, the simulations spanned multiple Reynolds and Péclet numbers, to fully represent highly advective and highly diffusive systems. The resulting mass transfer coefficients were compared with values predicted by common correlations and a good agreement was found at high Reynolds numbers, but less so at lower Reynolds number typical of cell operation, reinforcing the utility of the numerical approach. Dimensionless mass transfer correlations were determined for each material and each direction in terms of the Sherwood number. These correlations were analyzed with respect to each materials' permeability tensor. It was found that as the permeability of the system increases, the expected mass transfer coefficient decreases. Two general mass transfer correlations are presented, one correlation for isotropic fibrous media and the other for through-plane flow in planar fibrous materials such as electrospun media and carbon paper. The correlations are Sh = 0.879 Re 0.402 Sc 0.390 and Sh = 0.906 Re 0.432 Sc 0.432 respectively.
Defects in basal autophagy limit the nutrient supply from recycling of intracellular constituents. Despite our understanding of the prosurvival role of macroautophagy/autophagy, how nutrient deprivation, caused by compromised autophagy, affects oncogenic KRAS-driven tumor progression is poorly understood. Here, we demonstrate that conditional impairment of the autophagy gene Atg5 (atg5-KO) extends the survival of KRAS-driven tumor-bearing mice by 38%. atg5-KO tumors spread more slowly during late tumorigenesis, despite a faster onset. atg5-KO tumor cells displayed reduced mitochondrial function and increased mitochondrial fragmentation. Metabolite profiles indicated a deficiency in the nonessential amino acid asparagine despite a compensatory overexpression of ASNS (asparagine synthetase), key enzyme for de novo asparagine synthesis. Inhibition of either autophagy or ASNS reduced KRAS-driven tumor cell proliferation, migration, and invasion, which was rescued by asparagine supplementation or knockdown of MFF (mitochondrial fission factor). Finally, these observations were reflected in human cancer-derived data, linking ASNS overexpression with poor clinical outcome in multiple cancers. Together, our data document a widespread yet specific asparagine homeostasis control by autophagy and ASNS, highlighting the previously unrecognized role of autophagy in suppressing the metabolic barriers of low asparagine and excessive mitochondrial fragmentation to permit malignant KRAS-driven tumor progression.
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.