本文選題：可解釋性 + 人工智能技術(shù)　； 參考：《探索與爭(zhēng)鳴》2017年10期

【摘要】：正人工智能,從上世紀(jì)50年代到現(xiàn)在起起伏伏。我最早不喜歡神經(jīng)網(wǎng)絡(luò)的原因是在于它的不確定性,即最后輸出的結(jié)果是不確定的,今天深度學(xué)習(xí)卷積神經(jīng)網(wǎng)絡(luò)的不確定性依然存在。當(dāng)時(shí)從哲學(xué)意義上來講它是一種革新,所以大家很喜歡,到2005年又開始興起,起因是說它好像又非常好,把結(jié)果推廣到一些人類的競(jìng)賽上。其實(shí)在國(guó)際機(jī)器學(xué)習(xí)學(xué)術(shù)界依然是兩派,一部分還是看不起深度學(xué)習(xí),今天宣傳深度學(xué)習(xí)更多是工業(yè)背景。很多領(lǐng)域都開始嘗試深
[Abstract]:Positive artificial intelligence, from the 1950s to the present ups and downs. The reason why I didn't like neural network was that it was uncertain, that is, the result of final output was uncertain. Today, the uncertainty of neural network of deep learning convolution still exists. It was a philosophical innovation at the time, so people liked it, and it started to rise again in 2005, because it seemed to be very good again, and spread the results to some human race. In fact, in the international machine learning academic circles are still two schools, some still look down on in-depth learning, today's propaganda depth learning is more industrial background. A lot of areas are starting to go deep.
【作者單位】： 上海交通大學(xué)電子信息學(xué)院;
【分類號(hào)】：TP18
，

本文編號(hào)：1830493