240 发简信
IP属地:湖北
  • Resize,w 360,h 240
    《Neural Baby Talk》 CVPR2018

    与 《Knowing When to Look: Adaptive Attention via A Visual Sentinel for Im...

  • Resize,w 360,h 240
    Independently Recurrent Neural Network (IndRNN): Building A Longer and Deeper RNN

    Motivation plain的RNN存在梯度消失与爆炸问题,而LSTM/GRU虽然一定程度解决了问题,但是由于采用tanh或者sigmoid...