What defines an action like "kicking ball"? We argue that the true meaning of an action lies in the change or transformation an action brings to the environment. In this paper, we propose a novel representation for actions by modeling an action as a transformation which changes the state of the environment before the action happens (precondition) to the state after the action (effect). Motivated by recent advancements of video representation using deep learning, we design a Siamese network which models the action as a transformation on a high-level feature space. We show that our model gives improvements on standard action recognition datasets including UCF101 and HMDB51. More importantly, our approach is able to generalize beyond learned action categories and shows significant performance improvement on cross-category generalization on our new ACT dataset.

No abstract

In recent years, there has been a renewed interest in jointly modeling perception and action. At the core of this investigation is the idea of modeling affordances 1 . However, when it comes to predicting affordances, even the state of the art approaches still do not use any ConvNets. Why is that? Unlike semantic or 3D tasks, there still does not exist any large-scale dataset for affordances. In this paper, we tackle the challenge of creating one of the biggest dataset for learning affordances. We use seven sitcoms to extract a diverse set of scenes and how actors interact with different objects in the scenes. Our dataset consists of more than 10K scenes and 28K ways humans can interact with these 10K images. We also propose a two-step approach to predict affordances in a new scene. In the first step, given a location in the scene we classify which of the 30 pose classes is the likely affordance pose. Given the pose class and the scene, we then use a Variational Autoencoder (VAE) [23] to extract the scale and deformation of the pose. The VAE allows us to sample the distribution of possible poses at test time. Finally, we show the importance of large-scale data in learning a generalizable and robust model of affordances.

BackgroundMost state-of-the-art biomedical entity normalization systems, such as rule-based systems, merely rely on morphological information of entity mentions, but rarely consider their semantic information. In this paper, we introduce a novel convolutional neural network (CNN) architecture that regards biomedical entity normalization as a ranking problem and benefits from semantic information of biomedical entities.ResultsThe CNN-based ranking method first generates candidates using handcrafted rules, and then ranks the candidates according to their semantic information modeled by CNN as well as their morphological information. Experiments on two benchmark datasets for biomedical entity normalization show that our proposed CNN-based ranking method outperforms traditional rule-based method with state-of-the-art performance.ConclusionsWe propose a CNN architecture that regards biomedical entity normalization as a ranking problem. Comparison results show that semantic information is beneficial to biomedical entity normalization and can be well combined with morphological information in our CNN architecture for further improvement.

Large rivers are the main arteries for transportation of carbon to the ocean; yet, how hydrology and anthropogenic disturbances may change the composition and export of dissolved organic matter along large river continuums is largely unknown. The Yangtze River has a watershed area of 1.80 × 10 6 km 2 . It originates from the Qinghai-Tibet Plateau and flows 6300 km eastward through the center of China. We collected samples (n = 271) along the river continuum and analyzed weekly samples at the most downstream situated gauging station in 2017-2018 and gathered long-term (2006-2018) water quality data. We found higher gross domestic product, population density, and urban and agricultural land use downstream than upstream of the Three Gorges Dam, coinciding with higher dissolved organic carbon (DOC), UV absorption (a 254 ), specific ultraviolet absorbance (SUVA 254 ), parallel factor analysis-derived C1-C5, aliphatic compounds, and lower a 250 :a 365 and spectral slope (S 275-295 ). Chemical oxygen demand, humic-like C1-C2 and C6, and protein-like C4 and C7 increased, while dissolved oxygen and ammonium decreased with increasing discharge at most of the sites studied, including the intensively monitored downstream site. The annual DOC fluxes were ca. 1.5-1.8 Tg yr −1 , and 12-18% was biodegradable in a 28-d bio-incubation. Our results highlight that urbanization and stormwater periods enhanced the export of both terrestrial organic-rich substances and household effluents from nearshore residential areas. Our study emphasizes the continued need to protect the Yangtze River watershed as increased organic carbon loading or altered composition and bio-lability may change the ecosystem function and carbon cycling.Inland waters transport, transform, and store approximately 5 Pg of terrestrial carbon each year and are hot spots in the global cycling of dissolved organic matter (Drake et al. 2018). The associated dissolved organic carbon (DOC) fuels the net heterotrophy of fluvial and downstreamreceiving aquatic ecosystems (Tranvik et al. 2018) and contributes to the emission of greenhouse gases to the atmosphere. The composition and amount of dissolved organic matter are of special concern when river waters serve as sources of water supply due to the close link with its bio-lability and the chemical reactivity with heavy metals and micropollutants,