Yeah, this is basically stochastic dithering applied to numeric floating point quantization instead of image color quantization.
This makes me wonder whether you could apply different dithering approaches to numeric computations. You cannot use diffusion or similar mehods, because you don't have information about neighboring pixels/computations. Using low-discrepancy sequences might work to reduce stochastic noise, but it could also reintroduce bias for some computations.
alt Hacker News
Yeah, this is basically stochastic dithering applied to numeric floating point quantization instead of image color quantization.
This makes me wonder whether you could apply different dithering approaches to numeric computations. You cannot use diffusion or similar mehods, because you don't have information about neighboring pixels/computations. Using low-discrepancy sequences might work to reduce stochastic noise, but it could also reintroduce bias for some computations.