← Deep Learning learning pathDeep Learning · Module 4 of 10
Training with backpropagation
Learn weights with gradient descent and backpropagation.
📖 Key ideas
- Gradient descent moves weights in the direction that reduces loss.
- Backpropagation computes gradients layer by layer using the chain rule.
- Learning rate controls step size.
- Mini-batches process small groups of examples.
- Epoch: one pass over the full training data.
💡 Real-life example: Gradient descent is walking downhill in fog, feeling the slope to find the valley (lowest loss).
Gradient descentOptimisation by following the slope
BackpropagationComputes gradients through layers
Learning rateStep size of updates
EpochOne full pass over training data