Going Neural
摘要
A deep neural network can have millions of parameters. Where do they get their values from? Here is the trick: it starts with random values, calculates a first result starting from an observation (called propagation), and compares the result obtained with the expected result.