Title: From Neural Networks to Deep Learning

Abstract: Deep learning models, especially convolutional neural
networks, have achieved impressive successes in fields such as object
recognition and complex games. However, there are still many open
problems with the existing deep neural networks, namely, overfitting,
vanishing gradient, structure optimization, etc.  In this talk, the
research of neural netwoks will be summarized with four milestones in
its history. Through the analyses of the mechanism of each
representative neural network model, their advantages and
disadvantages will be discussed in depth.