人工智能,从上世纪50年代到现在起起伏伏。我最早不喜欢神经网络的原因是在于它的不确定性,即最后输出的结果是不确定的,今天深度学习卷积神经网络的不确定性依然存在。当时从哲学意义上来讲它是一种革新,所以大家很喜欢,到2005年又开始兴起,起因是说它好像又非常好,把结果推广到一些人类的竞赛上。其实在国际机器学习学术界依然是两派,一部分还是看不起深度学习,今天宣传深度学习更多是工业背景。很多领域都开始尝试深 Artificial intelligence, ups and downs from the 1950s to the present. My earliest reason I did not like neural networks was because of its uncertainty, that is, the result of the final output was indefinite. Today, the uncertainty of deep learning convolutional neural networks still exists. At that time it was an innovation in the philosophical sense, so we all enjoyed it and started to rise again in 2005. The reason is that it seems very good and extends the results to some human contests. In fact, in the international machine learning academic community is still two factions, some still look down on the depth of learning, today’s promotion of deep learning more is the industrial background. Many areas are beginning to try deep