This is SSRN - arXiv for social sciences. You can download the full text of the paper if you click.
The paper, I think, makes a more interesting claim than you give it credit for: those areas that can afford scholarly work tend to make use of free resources the most. Based on this work, it is possible that open access alone is not enough to give poorer universities, etc. the access to scholarly work.
It's not clear what "agenda" you think the paper has. It seems like an interesting piece of work.
How is it abuse? These resources have no marginal cost to produce and their creators generally want as much use as possible (think pre-prints, open access course material, ...). However, less well researched individuals (commonly found in poorer areas) are simply not aware of these resources.
If a data scientist at Google uses R instead of SAS (or some other proprietary statistical computing language) while a data scientist at a non-profit is forced to use SAS because of organizational constraints there is no abuse.
It's incredible how when its not China, you are always "its up to the community to decide". But when it comes to China, you go out of your way to freeze discussion. People like you are weak ass hypocrite pussies when it comes to China, you should be ashamed of yourself. Dan G you pussy ass communist appeaser, shame on you.
"Always" is a strong word. Usually it just means you noticed some things that you dislike [1]. The problem is that we're all far more likely to notice such cases and to weight them more strongly, so before long we've sample-biased ourselves to "always". The other side feels the opposite "always" [2]. Same mechanism in both cases. It always feels like the mods are against you, just as the refs are always against your team and you're always the one who gets the speeding ticket.
I feel like he nukes all my class war against the rich posts, but you have to realize I have personally accused of him fostering a white nationalist site for promoting assholes like you and also for hiding all my pro-communist comments.
So only one of us can be right, is dang a commie or nationalist?....or maybe everyone is complex and not entirely composed of a single ideology and also he was chosen to curate this website and that will ultimately anger both of us?
I said he was a communist appeaser. Curation is one thing. Let's not pretend like this motherfucker doesn't bend over backward and start singing about rules and civility when china is involved. That's fine, Its his website whatever, but let's not pretend he's not some sophist pussy like so much of the rest of america. He's like the dominant force of super sino simp on this site.
A few people have already published results of Geekbench 4 of the DTK and the performance is impressive already, specially considering that the DTK is using a 2-year old chip, in an emulated environment.
The chances that Intel will hit back is fairly small. In the last several years we have seen a gigantic brain-leak from the company, and is unlikely that they'll keep being reference for performance for much longer.
The demo images used on hific.github.io appear to be part of the datasets used to train the system. In another comment you say the trained model is 726MB.
The combined size of the training datasets appear to be about 8GB zipped.
Is the currently trained model usable on images that are not part of the training datasets? with output of similar quality and size?
Hi- no the images on the demo page are not part of the training set, they are only used to evaluate the method. Arbitrary images of natural scenes will look similar at these bitrates. We‘ll release trained models and a colab soon!
Thank you and my apologies. I should have read the pdf more carefully where the distinction between the training data and evaluation datasets is described
Our training set consists of a large set of high-resolution images collected from theInternet, which we downscale to a random size between 500 and 1000 pixels, and then randomlycrop to256×256. We evaluate on three diverse benchmark datasets collected independently ofour training set to demonstrate that our method generalizes beyond the training distribution: thewidely usedKodak[22] dataset (24 images), as well as theCLIC2020[43] testset (428 images), andDIV2K[2] validation set (100 images).
