Search for: All records

The eXtreme Multi-label Classification (XMC) problem seeks to find relevant labels from an exceptionally large label space. Most of the existing XMC learners focus on the extraction of semantic features from input query text. However, conventional XMC studies usually neglect the side information of instances and labels, which can be of use in many real-world applications such as recommendation systems and e-commerce product search. We propose Predicted Instance Neighborhood Aggregation (PINA), a data enhancement method for the general XMC problem that leverages beneficial side information. Unlike most existing XMC frameworks that treat labels and input instances as featureless indicators and independent entries, PINA extracts information from the label metadata and the correlations among training instances. Extensive experimental results demonstrate the consistent gain of PINA on various XMC tasks compared to the state-of-the-art methods: PINA offers a gain in accuracy compared to standard XR-Transformers on five public benchmark datasets. Moreover, PINA achieves a ∼ 5% gain in accuracy on the largest dataset LF-AmazonTitles-1.3M. Our implementation is publicly available https://github.com/amzn/pecos/ tree/mainline/examples/pina. 

The epitaxial growth of SrTiO 3 on Si(100) substrates that have been lithographically patterned to realize deposition-last, lithographically defined oxide devices on Si is explored. In contrast to traditional deposition-last techniques which create a physical hard mask on top of the substrate prior to epitaxial growth, a pseudomask is instead created by texturing the Si substrate surface itself. The Si is textured through a combination of reactive ion etching and wet-etching using a tetramethylammonium hydroxide solution. Desorbing the native SiO x at high temperatures prior to epitaxial growth in ultrahigh vacuum presents no complications as the patterned substrate is comprised entirely of Si. The inverted profile in which the epitaxial oxide device layer is above the textured pseudomask circumvents shadowing during deposition associated with conventional hard masks, thereby opening a pathway for highly scaled devices to be created. 

Surface cleaning using commercial disinfectants, which has recently increased during the coronavirus disease 2019 pandemic, can generate secondary indoor pollutants both in gas and aerosol phases. It can also affect indoor air quality and health, especially for workers repeatedly exposed to disinfectants. Here, we cleaned the floor of a mechanically ventilated office room using a commercial cleaner while concurrently measuring gas-phase precursors, oxidants, radicals, secondary oxidation products, and aerosols in real time; these were detected within minutes after cleaner application. During cleaning, indoor monoterpene concentrations exceeded outdoor concentrations by two orders of magnitude, increasing the rate of ozonolysis under low (<10 ppb) ozone levels. High number concentrations of freshly nucleated sub–10-nm particles (≥105 cm−3) resulted in respiratory tract deposited dose rates comparable to or exceeding that of inhalation of vehicle-associated aerosols.