Yann LeCun主讲！纽约大学《深度学习》2021课程全部放出，附slides与视频

来源：专知

本文附课程，建议阅读5分钟 Yann LeCun在纽约大学数据科学中心（CDS）主讲的《深度学习》2021年春季课程现已全部在线可看！

该课程自2021年春季开始由Yann LeCun与Alfredo Canziani等共同执教。

CDS发布了Yann LeCun的深度学习（DS-GA 1008）课程的所有材料，包括带英文字幕教学视频、书面讲义、课件以及带有PyTorch实现的可执行Jupyter Notebooks。

课程关注深度学习和表示学习的最新技术，重点关注有监督和无监督深度学习、嵌入方法、度量学习、卷积和循环网，以及在计算机视觉、自然语言理解和语音识别方面的应用。前提条件包括:DS-GA 1001数据科学入门或研究生水平的机器学习课程。

地址：

https://cds.nyu.edu/deep-learning/

资源

• YouTube视频：

  https://www.youtube.com/watch?v=mTtDfKgLm54

• 官方中文版讲义：

  https://atcold.github.io/pytorch-Deep-Learning/zh/

• 课件：

  https://github.com/Atcold/NYU-DLSP21

• GitHub：

  hhttps://atcold.github.io/NYU-DLSP21/

• Reddit论坛：

  https://www.reddit.com/r/NYU_DeepLearning/

授课老师：

目录内容：

Theme 1: Introduction

• History and resources 🎥 🖥

• Gradient descent and the backpropagation algorithm 🎥 🖥

• Neural nets inference 🎥 📓

• Modules and architectures 🎥

• Neural nets training 🎥 🖥 📓📓

• Homework 1: backprop

Theme 2: Parameters sharing

• Recurrent and convolutional nets 🎥 🖥 📝

• ConvNets in practice 🎥 🖥 📝

• Natural signals properties and the convolution 🎥 🖥 📓

• Recurrent neural networks, vanilla and gated (LSTM) 🎥 🖥 📓📓

• Homework 2: RNN & CNN

Theme 3: Energy based models, foundations

• Energy based models (I) 🎥 🖥

• Inference for LV-EBMs 🎥 🖥

• What are EBMs good for? 🎥

• Energy based models (II) 🎥 🖥 📝

• Training LV-EBMs 🎥 🖥

• Homework 3: structured prediction

Theme 4: Energy based models, advanced

• Energy based models (III) 🎥 🖥

• Unsup learning and autoencoders 🎥 🖥

• Energy based models (VI) 🎥 🖥

• From LV-EBM to target prop to (any) autoencoder 🎥 🖥

• Energy based models (V) 🎥 🖥

• AEs with PyTorch and GANs 🎥 🖥 📓📓

Theme 5: Associative memories

• Energy based models (V) 🎥 🖥

• Attention & transformer 🎥 🖥 📓

Theme 6: Graphs

• Graph transformer nets [A][B] 🎥 🖥

• Graph convolutional nets (I) [from last year] 🎥 🖥

• Graph convolutional nets (II) 🎥 🖥 📓

Theme 7: Control

1. Planning and control 🎥 🖥

2. The Truck Backer-Upper 🎥 🖥 📓

3. Prediction and Planning Under Uncertainty 🎥 🖥

Theme 8: Optimisation

• Optimisation (I) [from last year] 🎥 🖥

• Optimisation (II) 🎥 🖥 📝

Miscellaneous

• SSL for vision [A][B] 🎥 🖥

• Low resource machine translation [A][B] 🎥 🖥

• Lagrangian backprop, final project, and Q&A 🎥 🖥 📝

深度学习概要