Meta-Learning with Memory-Augmented Neural Networks

Timothy Lillicrap COUNTZERO@GOOGLE.COM Google DeepMind Abstract Despite recent breakthroughs in the applications ofdeepneuralnetworks,onesettingthatpresents a persistent challenge is that of “one-shot learn-ing.”Traditionalgradient-basednetworksrequire a lot of data to learn, often through extensive it-erative training. When new data is ...

  Team, Learning, Timothy, Team learning