In visual reasoning, the achievement of deep learning significantly improved the accuracy of results. Image features are primarily used as input to get answers. However, the image features are too redundant to learn accurate characterizations within a limited complexity and time. While in the process of human reasoning, abstract description of an image is usually to avoid irrelevant details. Inspired by this, a higher-level representation named semantic representation is introduced. In this paper, a detailed visual reasoning model is proposed. This new model contains an image understanding model based on semantic representation, feature extraction and process model refined with watershed and u-distance method, a feature vector learning model using pyramidal pooling and residual network, and a question understanding model combining problem embedding coding method and machine translation decoding method. The feature vector could better represent the whole image instead of overly focused on specific characteristics. The model using semantic representation as input verifies that more accurate results can be obtained by introducing a highlevel semantic representation. The result also shows that it is feasible and effective to introduce high-level and abstract forms of knowledge representation into deep learning tasks. This study lays a theoretical and experimental foundation for introducing different levels of knowledge representation into deep learning in the future.

In the field of visual reasoning, image features are widely used as the input of neural networks to get answers. However, image features are too redundant to learn accurate characterizations for regular networks. While in human reasoning, abstract description is usually constructed to avoid irrelevant details. Inspired by this, a higher-level representation named semantic representation is introduced in this paper to make visual reasoning more efficient. The idea of the Gram matrix used in the neural style transfer research is transferred here to build a relation matrix which enables the related information between objects to be better represented. The model using semantic representation as input outperforms the same model using image features as input which verifies that more accurate results can be obtained through the introduction of high-level semantic representation in the field of visual reasoning.

<p>Water cycle have prevailed on upper ocean salinity acting as the climate change fingerprint in the numerous observation and simulation works. Water mass in the Southern Ocean accounted for the increasing importance associated with the heat and salt exchanges between Subantarctic basins and tropical oceans. The circumpolar deep water (CDW), the most extensive water mass in the Southern Ocean, plays an indispensable role in the formation of Antarctic Bottom Water. In our study, the observed CTDs and reanalysis datasets are examined to figure out the recent salinity changes in the three basins around the Antarctica. Significant surface salinity anomalies occurred in the South Indian/Pacific sectors south of 60ºS since 2008, which are connected with the enhanced CDW incursion onto the Antarctic continental shelf. Saltier shelf water was found to expand northward from the Antarctica coast. Meanwhile, the freshening of Upper Circumpolar Deep Water(UCDW), salting and submergence of Subantarctic Mode Water(SAMW) were also clearly observed. The modified vertical salinity structures contributed to the deepen mixed layer and enhanced intermediate stratification between SAMW and UCDW. Their transport of salinity flux attributed to the upper ocean processes responding to the recent atmospheric circulation anomalies, such as the Antarctic Oscillation and Indian Ocean Dipole. The phenomena of SAMW and UCDW salinity anomalies illustrated the contemporaneous changes of the subtropical and polar oceans, which reflected the meridional circulation fluctuation. Salinity changes in upper southern ocean (< 2000m) revealed the influence of global water cycle changes, from the Antarctic to the tropical ocean, by delivering anomalies from high- and middle-latitudes to low-latitudes oceans.</p>