Paul Hand Courses

Continual Learning and Catastrophic Forgetting

1 rating

Explore how deep neural networks can learn new tasks without catastrophically forgetting previous knowledge through regularization, dynamic architectures, and generative replay.

• YouTube
• 42 minutes
• Self-Paced
• Free Video

Variational Autoencoders

Explore how variational autoencoders generate data through latent codes, balancing reconstruction accuracy and distribution regularization with the variational lower bound.

• YouTube
• 43 minutes
• Self-Paced
• Free Video

Gradient Descent and Stochastic Gradient Descent

Analyze how learning rates shape the convergence, divergence, and optimization behavior of gradient descent and stochastic gradient descent in deep neural networks.

• YouTube
• 57 minutes
• Self-Paced
• Free Video

Adversarial Examples for Deep Neural Networks

Explores how small, human-imperceptible perturbations fool deep neural networks, covering white- and black-box attacks, physical-world examples, and adversarial training defenses.

• YouTube
• 44 minutes
• Self-Paced
• Free Video

Neural Network Architectures for Images

Survey neural network architectures for image classification, segmentation, restoration, super-resolution, and reconstruction, including convolutional, residual, encoder-decoder, and autoencoder designs.

• YouTube
• 47 minutes
• Self-Paced
• Free Video

Supervised Machine Learning Review

Review supervised learning through loss functions, bias-variance tradeoffs, and the surprising double-descent behavior of highly overparameterized neural networks.

• YouTube
• 47 minutes
• Self-Paced
• Free Video

Unlearned Neural Networks as Image Priors for Inverse Problems

Explore unlearned neural networks as image priors for solving imaging inverse problems without training data.

• YouTube
• 50 minutes
• Self-Paced
• Free Video

Theory of GANs for Compressed Sensing

A theoretical lecture on using GAN priors to recover natural images from underdetermined compressed-sensing measurements.

• YouTube
• 57 minutes
• Self-Paced
• Free Video