From Dev.to

Why did deep networks stop learning?

Blame travels backward When a neural network makes a mistake, training works out how much each dial (each weight) was to blame. That blame starts at the answer and travels backward, layer by layer,...

Read the full report at Dev.to ↗