Thank you for raising the interesting question on the conditions for asymptotic
–Neural Information Processing Systems
This is achieved e.g. if a constant fraction of all samples lies on the point Theorem 3.3 by reformulating lines 190-191 as follows: "Furthermore, consider an infinite data stream of observations ". Making Theorem 3.3 quantitative as suggested by Reviewer #2 Although unbounded, they grow slow enough to allow the proof of Theorem 3.3 such that the main We will add a brief discussion on this in the updated paper. Reviewer #1 pointed out, that Assumption 3.1. Therefore, Assumption 3.1 is valid for our experimental setup. We will include the given reasoning in the updated paper.
Neural Information Processing Systems
Aug-20-2025, 11:22:21 GMT
- Technology: