Every Model Learned by Gradient Descent Is Approximately a Kernel Machine (arxiv.org)
recommended by 1 user | 2020 | 406 points by scottlocklin | machine learning | original thread |
One user recommended this thread. What they said:
6gvONxR4sf7o Feb 8, 2021 | on: Gradient Descent Models Are Kernel Machines | context

Discussion of the actual paper here:

https://news.ycombinator.com/item?id=25314830

It's a really neat one for people who care about what's going on under the hood, but not immediately applicable to the more applied folks. I saw some good quotes at the time to the tune of "I can't wait to see the papers citing this one in a year or two."