Learn Deep Learning with this Free Course from Yann LeCun
Here is a freelyavailable NYU course on deep learning to check out from Yann LeCun and Alfredo Canziani, including videos, slides, and other helpful resources.
It goes without saying that there are a nearly unlimited number of resources available for learning deep learning from scratch. This applies at least equally to freelyavailable resources, it not more so than paid. A course which has been on the community's radar recently and being shared widely across social media is the aptly titled Deep Learning course from the NYU Center for Data Science, taught by Yann LeCun & Alfredo Canziani.
Here is an overview of the course, directly from its website:
This course concerns the latest techniques in deep learning and representation learning, focusing on supervised and unsupervised deep learning, embedding methods, metric learning, convolutional and recurrent nets, with applications to computer vision, natural language understanding, and speech recognition. The prerequisites include: DSGA 1001 Intro to Data Science or a graduatelevel machine learning course.
The course provides a balanced introduction to a wide variety of deep learning concepts. A nonexhaustive list of course lecture and practicum topics include:
 SGD and backprop
 Backprop in practice
 NN training
 Parameter transformation
 CNN, autograd
 CNN applications
 RNNs and attention
 Training RNNs
 Autoencoders
 Contrastive methods
 Regularised latent
 Training GANs
 Activations
 Losses
 DL for NLP
 Attention & transformer
 Structured Prediction
 Regularisation and Bayesian
 Inference for LatentVariable EBMs
 Training LatentVariable EBMs
The website also acts as an online text of sorts, with structured HTML course notes compiled by lecture and section readable inplace.
Course material also includes executable Jupyter Notebooks with PyTorch implementations of various course concepts, which can be accessed directly here.
The course's YouTube lectures playlist can be directly accessed here.
Impressively, translations of the course materials are presented in: English, Arabic, Spanish, Italian, Japanese, Korean, Turkish, Chinese, French, Persian, Russian, with plans to add Portuguese, Bengali, and Vietnamese. These translations have been facilitated by more than 470 volunteers from 17 time zones around the world.
The opportunity to learn deep learning from a renowned deep learning researcher and expert from a topnotch university (not to mention without cost) represent a fantastic opportunity for all. The effectiveness of Canziani's approachable and explanatory videos are also not to be overlooked, being substantial highlights of the course materials. I would encourage everyone to have a look for themselves to see if this course is for them.
Related:
 10 Days With "Deep Learning for Coders"
 Free From MIT: Intro to Computer Science and Programming in Python
 Free From MIT: Intro to Computational Thinking and Data Science
Top Stories Past 30 Days

