Deep Learning Foundations

Introduction to Deep Learning

Deep learning uses neural networks with several layers to learn useful representations from data. Each layer transforms the values it receives, and training adjusts the weights in those transformations to reduce prediction error.

Begin with one input, such as the temperature of a machine. A neuron multiplies that value by a weight, adds a bias, and passes the result through an activation function. Its output can then become the input to another layer.

A single linear calculation can draw only a straight decision boundary. Stacking layers with nonlinear activation functions allows the network to represent curved and more complicated relationships.

During training, the network makes a prediction and the loss measures how far that prediction is from the desired answer. A derivative tells us how a small change to one weight would change that loss.

Backpropagation applies the chain rule from the output layer toward the input layer so every trainable weight receives a gradient. The optimizer then uses those gradients to adjust the weights before the next batch.