topic technology ★★ seed
Deep Learning
Neural networks with multiple layers that learn hierarchical representations.
Deep learning stacks layers of artificial neurons to learn representations at increasing levels of abstraction — edges become textures become objects become scenes. The key insight is that features do not need to be hand-engineered; given enough data and compute, gradient descent discovers them. Backpropagation makes this tractable by efficiently computing gradients through the chain rule. The revolution began with convolutional networks for vision (AlexNet, 2012), expanded to sequence modeling with recurrent networks, and was transformed again by the attention mechanism.
#deep-learning
#neural-networks
#representation-learning