Search for: All records

Although self-/un-supervised methods have led to rapid progress in visual representation learning, these methods generally treat objects and scenes using the same lens. In this paper, we focus on learning representations for objects and scenes that preserve the structure among them. Motivated by the observation that visually similar objects are close in the representation space, we argue that the scenes and objects should instead follow a hierarchical structure based on their compositionality. To exploit such a structure, we propose a contrastive learning framework where a Euclidean loss is used to learn object representations and a hyperbolic loss is used to encourage representations of scenes to lie close to representations of their constituent objects in a hyperbolic space. This novel hyperbolic objective encourages the scene-object hypernymy among the representations by optimizing the magnitude of their norms. We show that when pretraining on the COCO and OpenImages datasets, the hyperbolic loss improves downstream performance of several baselines across multiple datasets and tasks, including image classification, object detection, and semantic segmentation. We also show that the properties of the learned representations allow us to solve various vision tasks that involve the interaction between scenes and objects in a zero-shot fashion. 

Syntheses of Rh complexes of the phosphine-amido-silane SiNP ligand are reported. The reaction of the parent (SiNP)H ligand (4) with 0.5 equiv. [(COE)RhCl] 2 (COE = cis -cyclooctene) in the presence of NaN(SiME 3 ) 2 resulted in the formation of (SiNP)Rh(COE) (5). Compound 5 was converted to a series of (SiNP)Rh(P(OR) 3 ) complexes 6–10 (R = Ph, i Pr, n Bu, Et, or Me) by treatment with the corresponding phosphite. NMR and XRD structural data, as well as the DFT computational analysis indicate that compounds 5–10 are divided into two structural Types ( A and B ), differing in the nature of the interaction of the Si–H bond of the SiNP ligand with Rh. 

State-of-the-art subspace clustering methods are based on the self-expressive model, which represents each data point as a linear combination of other data points. However, such methods are designed for a finite sample dataset and lack the ability to generalize to out-of-sample data. Moreover, since the number of self-expressive coefficients grows quadratically with the number of data points, their ability to handle large-scale datasets is often limited. In this paper, we propose a novel framework for subspace clustering, termed Self-Expressive Network (SENet), which employs a properly designed neural network to learn a self-expressive representation of the data. We show that our SENet can not only learn the self-expressive coefficients with desired properties on the training data, but also handle out-of-sample data. Besides, we show that SENet can also be leveraged to perform subspace clustering on large-scale datasets. Extensive experiments conducted on synthetic data and real world benchmark data validate the effectiveness of the proposed method. In particular, SENet yields highly competitive performance on MNIST, Fashion MNIST and Extended MNIST and state-of-the-art performance on CIFAR-10. 

ABSTRACT Image simulations are essential tools for preparing and validating the analysis of current and future wide-field optical surveys. However, the galaxy models used as the basis for these simulations are typically limited to simple parametric light profiles, or use a fairly limited amount of available space-based data. In this work, we propose a methodology based on deep generative models to create complex models of galaxy morphologies that may meet the image simulation needs of upcoming surveys. We address the technical challenges associated with learning this morphology model from noisy and point spread function (PSF)-convolved images by building a hybrid Deep Learning/physical Bayesian hierarchical model for observed images, explicitly accounting for the PSF and noise properties. The generative model is further made conditional on physical galaxy parameters, to allow for sampling new light profiles from specific galaxy populations. We demonstrate our ability to train and sample from such a model on galaxy postage stamps from the HST/ACS COSMOS survey, and validate the quality of the model using a range of second- and higher order morphology statistics. Using this set of statistics, we demonstrate significantly more realistic morphologies using these deep generative models compared to conventional parametric models. To help make these generative models practical tools for the community, we introduce galsim-hub, a community-driven repository of generative models, and a framework for incorporating generative models within the galsim image simulation software. 

Influenced by weather, the estuaries and bays often exhibit recurring oscillations in flow and water level similar to astronomical tides. The weather impact however is less regular than tides and more difficult to predict. The spectrum of weather induced motions in estuaries and bays is mostly at the low-frequency end with time scales longer than those of diurnal tides. The repeated weather impact produces meteorological tide: the recurring flood and ebb and flushing of the estuaries and bays but at lower frequencies than those of tides. The variation in weather conditions is quasi-periodic and of large scale nature (~1000−3000 km) because of the alternating low- and high- atmospheric pressure systems of extra-tropical cyclones and anti-cyclones and associated fronts. By examining 40 years of data between Jan. 1, 1977 and Dec. 31, 2016, we identified 1648 frontal events (averaging ~41.2±4.7 per year) influencing the northern Gulf of Mexico for time periods in the spring, fall and winter. The late spring and summer months (May, Jun, July, and August) were not included in the calculation because of much weaker activities involving synoptic weather systems with fronts during these months. It is found that the number of frontal events reached the maximum in Jan. and Dec. while the minimum occurred in April and Sept. It is found that there is an increasing trend of number of fronts over the 40-year period. Our data show that the low pass filtered water level, velocity, and vorticity (velocity shear) all vary in response to the weather and appear as the meteorological tide. The particle excursions of meteorological tides are much larger than those from the astronomical tides. In addition, the irregular nature of the meteorological tide makes the inward flux and outward flux asymmetric in general and thus it has a significant implication to dispersion and transport of waterborne materials. A scaling analysis shows that the meteorological tide generally reaches quasi-steady state; and as a result, a regression model is established which can be very useful for predicting the weather produced quasi-periodic motions.