mirker

26 ms·

241.
by mirker 5y ago
I see. Some of these, like ipfs, are more general purpose systems (basically communication protocols) than I was thinking of.
242.
by mirker 5y ago
There are 3 ways to allocate cloud resources: 1. Use GUI 2. Use Boto library with Python 3. Use devops tools like Terraform As you go from 1 to 3, the programmability increases. 3 is additionally idempotent. As others mention, 3 has competi
243.
by mirker 5y ago
Can you elaborate on what requirements these blob systems should have? My understanding is object stores are typically “flat” by design to scale well (in contrast to a tree structure found in filenames). For content addressing, are people u
244.
by mirker 5y ago
Gradients and convex analysis pop up in optimization. My understanding is controls uses analysis as well (at least for understanding theory).
245.
by mirker 5y ago
More like linear algebra, real analysis, and probability for the areas you listed.
246.
by mirker 5y ago
(For sufficiently small data, because the $/GB difference is still there).
247.
by mirker 5y ago
You’re ultimately at the mercy of the SSD software, which was written to minimize development costs for a limited-product-lifetime embedded device. Most of these quirks are undefined behavior, and you won’t ever see the source code.
248.
by mirker 5y ago
Right. Listing 10 papers with “ok” impact should be viewed as 10 “bad” papers and not 10 “good” papers. It’s extremely unlikely that a 50 page CV is demonstrating real value throughout. It’s the academic equivalent of putting 20 Javascript
249.
by mirker 5y ago
Top N papers is already a metric used for professorship slots, so Max or other functions are me simplifying the situation. I agree though that citations are proxies for goodness and are ultimately poor indications of work. For example, a pa
250.
by mirker 5y ago
One problem is the incentives are biased toward “ok” papers. h-index should be replaced with a more competitive index (e.g., Max over citations). The publication hyper-competitiveness is due to publication spam, which boosts citation and pu
251.
by mirker 5y ago
Even if insider information gives you a small percentage advantage over uninformed trading, that still increases expected gains unfairly. A hedge fund would probably be quite happy with a 5-10% signal over random chance.
252.
by mirker 5y ago
You can have a log of N register “transactions”, where N is large enough to hide core-to-core communication. If any of N transactions fail due to mismatch between cores, you roll back and throw exceptions. A lot of this logic is already in
253.
by mirker 5y ago
I’m familiar with the site but I didn’t realize it was used as a primary source. If so, then I think this is a good example.
254.
by mirker 5y ago
It’s not necessary to serialize the full execution to detect errors. On an out of order processor, there is already buffering of results happening that is eventually serialized to visible state. To check errors, you could just have one buff
255.
by mirker 5y ago
In my opinion, data should be mirrored on a torrent in addition to institutional servers (which can also provide checksums). Torrents offload the bandwidth problems from the institute to users and stay up if people use the data. But that pr
256.
by mirker 5y ago
Yep, it’s more meaningful for someone to go out of their way to meet up with you (when there’s only friendship to gain). Also prevents mixing personal life with work.
257.
by mirker 5y ago
What is the typical range of mean time to attack? Seconds? Is it fair to model the attack as random coin flips (i.e., following a geometric distribution)?
258.
by mirker 5y ago
And each level likely requires an exponentially larger dataset.
259.
by mirker 5y ago
My understanding is the title is “fastest”, which is 3.8 exaflops of sparse fp16. I assume these are theoretical peak or “benchmark” peak performance. Replicating their hardware in the cloud would get you the same title until a concrete wor
260.
by mirker 5y ago
There’s more to performance than peak theoretical performance. The architectures that stick around have tended to be ones with more software support (e.g., x86). If you mean network bandwidth, the level of batching controls the bandwidth bo