Lecture 1History and resources

Prev1 / 33Next

Lecture 1: History and resources

History and resources

Lecture 1 Lab: Gradient descent and the backpropagation algorithm

Lecture 2: Neural nets: rotation and squashing

Lecture 2 Lab: Modules and architectures

Lecture 3: Tools, classification with neural nets, PyTorch implementation

Lecture 3 Lab: Parameter sharing: recurrent and convolutional nets

Lecture 4 Lab: ConvNet in practice

Lecture 4.1: Natural signals properties and the convolution

Lecture 4.2: Recurrent neural networks, vanilla and gated (LSTM)

Lecture 5 Lab: Joint embedding method and latent variable energy based models (LV-EBMs)

Lecture 5.1: Latent Variable Energy Based Models (LV-EBMs), inference

Lecture 5.2: But what are these EBMs used for?

Lecture 6 Lab: Latent variable EBMs for structured prediction

Lecture 6: Latent Variable Energy Based Models (LV-EBMs), training

Lecture 7 Lab: PCA, AE, K-means, Gaussian mixture model, sparse coding, and intuitive VAE

Lecture 7: Unsupervised learning: autoencoding the targets

Lecture 8 Lab: Self-supervised learning and variational inference

Lecture 8: From LV-EBM to target prop to autoencoder

Lecture 9 Lab: Differentiable associative memories, attention, and transformers

Lecture 9: AE, DAE, VAE, and GANs with PyTorch

Lecture 10 Lab: Self-supervised learning in computer vision

Lecture 10: Self / cross, hard / soft attention and the Transformer

Lecture 11 Lab: Speech recognition and Graph Transformer Networks

Lecture 11: Graph Convolutional Networks (GCNs)

Lecture 12 Lab: Low resource machine translation

Lecture 12: Planning and control

Lecture 13 Lab: Optimisation for Deep Learning

Lecture 13: The Truck Backer-Upper

Lecture 14 Lab: Lagrangian backpropagation, final project winners, and Q&A session

Lecture 14: Prediction and Planning Under Uncertainty

Seminar: Energy-Based Self-Supervised Learning

Practice 9: Contrastive joint embedding methods for self-supervised learning

Practice 10: Non-contrastive joint embedding methods for self-supervised learning