20160203_마인즈랩_딥러닝세미나_02 deep and wide analytics 최대우교수님

The document discusses the history and development of neural networks. It covers early work from 1943-1986 establishing basic neural network concepts. These include McCulloch and Pitts' binary input/output model in 1943 and Rosenblatt's perceptron in 1957. The document then discusses later developments, including Kolmogorov establishing neural networks as universal approximators in 1963. It describes a "dark age" for neural networks from the 1960s-1980s due to issues like overfitting. Later sections cover how developments like autoencoders, dropout, and GPUs helped address issues and enabled modern deep learning applications. The document concludes by discussing future areas like using more data and models to reduce optimism in predictions, and applications of reinforcement learning.