rsfern

13 ms·

121.
by rsfern 2y ago
They’re not writing the whole paper this way, I think they have the student write a complete first draft and then co-edit one paragraph at a time. Hopefully the student improves their editing skills in less than a full papers worth of co-ed
122.
by rsfern 2y ago
Thanks for the link, that is a really nice paper! I found it really refreshing that they report a bunch of stuff they tried that didn’t work in a way that clarifies the problem and leads to a lot of insight into the strengths and limitation
123.
by rsfern 2y ago
This new technique is done in a scanning electron microscope, which has some advantages - I’d expect you could get better temporal resolution since this scans an electron beam vs scanning a solid probe tip for magnetic force microscopy, and
124.
by rsfern 2y ago
Yeah that sounds awesome to me. It addresses one of the shortcomings of the conventional abstract style which is that nothing is cross referenced - most journals still disallow citations in the abstract A feature like this would be great fo
125.
by rsfern 2y ago
Sure, thanks for building this cool site! Thinking about the hyperlink thing more, I think I would revise my suggestion to avoid undefined acronyms - I think you could explicitly encourage them, but only if they are tagged and defined and l
126.
by rsfern 2y ago
I feel like this is basically already what people who promote their papers on twitter are doing. Other than Twitter or LinkedIn or something (I don’t really use either) I’m not sure there’s really a place for researchers to actually post su
127.
by rsfern 2y ago
I think it’s an interesting idea. It could use a documentation page with some clear examples and some guidelines. Don’t use undefined acronyms, try to make it self contained, etc. For example, the tldr(s) “attention is all you need” don’t a
128.
by rsfern 2y ago
They grew CaF2 crystals with a small amount of thorium, which take the place of some of the Ca atoms in the crystal lattice. There’s an illustration in fig 1 of the preprint of what the substitutional defects look like Like the other poster
129.
by rsfern 2y ago
Not sure if the wrapper you’re talking about is your own custom code, but I really like using einops lately. It’s got similar axis naming capabilities and it dispatches to both numpy and pytorch http://einops.rocks/
130.
by rsfern 2y ago
Power electronics, maybe? https://ece.illinois.edu/newsroom/63402
131.
AI and Open Government Data Assets Request for Information(federalregister.gov)
2 points by rsfern 2y ago|0 comments
132.
by rsfern 2y ago
Sorry - I didn’t mean to come off as discouraging, that’s part of why I linked the BEAR paper! I agree in relative terms, I’d think this cellular structure should have better compressive strength than a different structure made of the same
133.
by rsfern 2y ago
It’s not that kind of 3D printing, it’s laser powder bed fusion [0]. I don’t know a lot about FDM but I’d be kind of surprised if you could do Ti-64 and get the same mechanical strength (but I’d be interested to know if I’m off base) becaus
134.
by rsfern 2y ago
To be honest I don’t know too much about the biomedical side, I’m probably jumping to conclusions on what the materials research directions are there - originally I was just going to joke about how you might hope it doesn’t hit its melting
135.
by rsfern 2y ago
It’s a titanium alloy (Ti-6-4) that’s commonly used in medical implants, and it probably melts at pretty high temp - the wrought form is used in turbine engine compression disks This is like a 3D printed metal scaffold, in the paper they’re
136.
by rsfern 2y ago
I don’t think anything needs to be shut down necessarily, I’m just suggesting reasons why a reasonable ethics board might be hesitant to green light such a hackathon if it’s not clear the organizers have done their due diligence on safety I
137.
by rsfern 2y ago
How do you make sure the participants are well informed? What if an idea suggested by a model turns out to be dangerous to implement, but nobody at the hackathon has quite the relevant experience to notice?
138.
by rsfern 2y ago
There’s some interesting work replacing scaled dot product attention and position embeddings with fixed format MLPs [0] - so I tend to lean towards thinking of classic transformers as having a reasonable enough inductive bias and the scalab
139.
by rsfern 3y ago
To be fair, they do call it a “proof of concept study” in the press release
140.
by rsfern 3y ago
I don’t think the magnetic particles interact with the glucose, they’re just a magnetic transducer basically. The main measurement device is a hydrogel that reacts with the glucose and swells, and they measure how far the magnetic particles
141.
by rsfern 3y ago
I think it’s just that the time/frequency analogy is so strongly rooted in signals processing applications In the physics and crystallography literature the Fourier transform is framed in terms of real space and reciprocal space, which
142.
by rsfern 3y ago
More physical ML force fields is a super interesting topic that I feel like blurs the line between ML and actually just doing physics. My favorite topic lately is parametrizing tight binding models with neural nets, which hopefully would le
143.
by rsfern 3y ago
NIST has a similar setup. There’s an exemption for e.g. summer students who are issued temporary non-PIV badges, but they’re issued a yubikey that’s required to access the network from off campus
144.
by rsfern 3y ago
A counter perspective: it’s a good thing these ideas make it to the surface! Clearly someone thought it was a good enough idea to try, now others can have better data before trying that same rabbit hole The sad part about the whole situatio
145.
by rsfern 3y ago
It’s also useful for getting hot takes by reviewers who didn’t read the paper carefully. Though even in those cases sometimes the author responses can be pretty illuminating
146.
by rsfern 3y ago
it would be great if you could provide a bit more detail to your criticism, especially pointers to relevant literature you think they missed Frankly I don’t really see how some discrete optimization thing solves the problem this paper is a
147.
by rsfern 3y ago
You could still have assigned reviewers like in a conventional journal, which maybe if done right could stimulate broader discussion JOSS has a review process like this, but I don’t think I’ve heard of any journals focused on primary scient
148.
by rsfern 3y ago
Attention (for quality work or not) is the coin of the realm, so why would you expect researchers not to optimize for flashy clickbait over time consuming rigorous self-reflecting work? People have been identifying the incentive structure
149.
by rsfern 3y ago
It’s not just grain size strengthening, it seems the surprising magic is a particle strengthening effect from dislocation pinning on nanoscale porosity
150.
by rsfern 3y ago
In principle, I think so, but there a lot of technical barriers to learning how to actually make the structures you’d need to do it. (I haven’t read the actual paper because of the paywall) From my understanding, there are (at least) 3 fact

More ›