Search

Interpreting how does deep neural networks (DNNs) make predictions is a vital field in artificial intelligence, which hinders wide applications of DNNs. Visualization of learned representations helps we humans understand the vision of DNNs. In this work, visualized images that can activate the neural network to the target classes are generated by back-propagation method. Here, rotation and scaling operations are applied to introduce the transformation invariance in the image generating process, which we find a significant improvement on visualization effect. Finally, we show some cases that such method can help us to gain insight into neural networks., 15 pages, 7 figures

The Ni-catalyzed asymmetric hydrogenation of challenging β-acylamino nitroolefins was achieved under mild conditions, affording β-acylamino nitroalkanes in excellent yields and with high enantioselectivities., An efficient approach for synthesizing chiral β-amino nitroalkanes has been developed via the Ni-catalyzed asymmetric hydrogenation of challenging β-amino nitroolefins under mild conditions, affording the desired products in excellent yields and with high enantioselectivities. This protocol had good compatibility with the wide substrate scope and a range of functional groups. The synthesis of chiral β-amino nitroalkanes on a gram scale has also been achieved. In addition, the reaction mechanism was elucidated using a combined experimental and computational study, and it involved acetate-assisted heterolytic H2 cleavage followed by 1,4-hydride addition and protonation to achieve the nitroalkanes.