#imagenet

Question 1: Single choice

Multi-layer neural networks had been known for decades, yet around 2012 deep networks such as AlexNet suddenly beat other image-recognition methods by a wide margin. What changed?

Large labelled datasets such as ImageNet and fast GPU computing became available1
Backpropagation was invented, so deep networks could be trained for the first time2
Expert systems began to supply hand-written rules to the networks3
Networks were made shallower, which made them much easier to train4

Timed test