Fetching the paper…

Implicit Acceleration and Feature Learning in Infinitely Wide Neural Networks with Bottlenecks · Around