Every Model Learned by Gradient Descent Is Approximately a Kernel Machine

406 points by scottlocklin

Read the article

Discussion on Hacker News