Comment on Every Model Learned by Gradient Descent Is Approximately a Kernel MachineparentComments−Straw5yOh, I don't know, it very well might! However, the brain's physical structure doesn't look conducive to 4096x4096 matmuls!
Comments
Oh, I don't know, it very well might! However, the brain's physical structure doesn't look conducive to 4096x4096 matmuls!