Explore energy-based approaches to representation learning in this 40-minute lecture by Yann LeCun from NYU and Facebook AI. Delve into topics such as self-supervised learning, video prediction, energy functions, latent variable models, autoencoders, and sparse modeling. Gain insights on how humans and animals learn, the right framework for building predictors, and the applications of energy-based models in high-dimensional continuous spaces. Discover the connections between sparse coding, sparse autoencoders, and linear decoders in convolutional models. Learn about the implications of these approaches for model predictive control and advanced AI systems.
Energy-Based Approaches to Representation Learning - Yann LeCun
Institute for Advanced Study via YouTube
Overview
Syllabus
Intro
How do humans and animals learn
Selfsupervised running
Video prediction
Building a predictor
The right framework
Thin plates
Energy functions
Energybased model
Latent variable model
Autoencoders
probabilistic modeling
highdimensional continuous spaces
sparse coding
sparse modeling
sparse autoencoder
linear decoder
convolutional
model predictive control
Taught by
Institute for Advanced Study
Related Courses
-
Contrastive Methods and Regularised Latent Variable Models
-
PCA, AE, K-Means, Gaussian Mixture Model, Sparse Coding, and Intuitive VAE
-
Joint Embedding Method and Latent Variable Energy Based Models
-
Energy Based Models and Self-Supervised Learning
-
Self-Supervised Learning - The Dark Matter of Intelligence
-
Unsupervised Representation Learning
