Explore a comprehensive lecture on optimization techniques in deep learning, covering gradient descent, stochastic gradient descent (SGD), and momentum updates. Delve into adaptive methods like RMSprop and ADAM, and understand the impact of normalization layers on neural network training. Learn about the intuition behind these concepts, their performance comparisons, and their effects on convergence. Discover a real-world application of neural networks in accelerating MRI scans, demonstrating the practical implications of optimization in industry.
Overview
Syllabus
– Week 5 – Lecture
– Gradient Descent
– Stochastic Gradient Descent
– Momentum
– Adaptive Methods
– Normalization Layers
– The Death of Optimization
Taught by
Alfredo Canziani
