It seems part of the problem is that only health insurance premiums that exceed 10% of your income are tax-deductible. Insurance through employers comes fully from pre-tax income. This amounts to a significant penalty to those obtaining insurance separate from employment.
Does anyone know why this is the case?
(Not an expert on tax law, so correct me if I’m wrong on this.)
A beautiful related result is the Crofton Formula [0], which says that you can measure the length of any curve by appropriately counting the number of straight lines that intersect it.
(My co-authors and I discovered a generalization of the formula that also holds in curved space [1].)
Author here— Yes, this is a great question! I think there's a lot more interesting work to be done here. It would be great if we could understand e.g. why layer-wise learning rates help with large-batch training of ImageNet (https://arxiv.org/abs/1708.03888), and maybe the per-component noise scale has something to do with it.
Another author here– I'll add to Jared's comment above that for long-running experiments (like the ones our Dota team runs), it can be useful to track this statistic in real time to see whether or not it would be useful to scale up the experiment.
One of the authors here. Thanks for the comment! Yes, we mention this work and number of others in the blogpost and paper. This isn't the first (or the last) paper on the topic but I think we've clarified the large-batch training situation significantly by connecting gradient noise directly to the speed of training, and by measuring it systematically on a bunch of ML tasks and characterizing its behavior.
I often wonder how much damage the term "affordable housing" has done to the goal of true housing affordability.
Politicians can pretend to be working on improving housing prices, while in reality doing nothing of the sort. As far as I can tell, "affordable housing" requirements produce only false hope for lower-income residents and increased market rates.
True, but there are some kinds of data that people are still uncomfortable sending to Google. Medical data is considered especially private, and aggregation of medical data is a huge obstacle to improving treatment using ML. I think this could be really huge in that space.
When I bring the screen closer to my face, the area where I can detect the movement becomes larger. Far away from my face, I can see only a few spinning. If this area corresponds to my fovea, wouldn't I expect the opposite? Perhaps larger spinners can be detected father away from the fovea?
Since we experience this distortion for anything we look at, aren't our brains equipped to handle it? Can we really call it a distortion if that is true?
Does anyone know why this is the case?
(Not an expert on tax law, so correct me if I’m wrong on this.)