整理 | 周翔 作為深度學(xué)習(xí)的奠基人之一,Yann LeCun 的一言一行都頗受關(guān)注。近日,Yann LeCun 在 CCN 2017 (認(rèn)知計(jì)算神經(jīng)科學(xué)大會(huì))上發(fā)表了主題為“How does the Brain learn so much so quickly”的演講,內(nèi)容詳實(shí),非常有啟發(fā)性。 該演講主要分為以下幾個(gè)部分: 1)Obstacles to AI(通往 AI 的阻礙)
2)The Architecture of an Intelligent System(智能系統(tǒng)的架構(gòu)) 3)Predictive Models In Questioning-Answering and Dialog(問答和對(duì)話中的預(yù)測(cè)模型) 4)Predictive Models with Uncertainty Adversarial Training(具有不確定性對(duì)抗訓(xùn)練的預(yù)測(cè)模型) 5)Video Prediction of semantic Segmentation(基于語義分割的視頻預(yù)測(cè)) 6)Semi-Supervised Learning (adversarially)(半監(jiān)督學(xué)習(xí)) 7)Fader Networks: Disentangling factors of variation to Parameterize the Image Manifold (adversarially) 8)How do we 'design' objective functions? We learn them 以下是AI科技大本營(yíng)整理的完整PPT內(nèi)容和視頻播放地址: B站地址: https://www.bilibili.com/video/av15938572/ Youtube地址: https://www./watch?v=cWzi38-vDbE |
|