Unlike PulseAudio, dbus, and other userland components, it is perfectly possible to have a Linux system which works with most software without systemd. runit and OpenRC are two of the most popular init alternatives, which are just inits, and nothing else, unlike systemd. You might argue you have to use logind and udev, but that has been spun off into elogind and eudev. There is also seatd as an alternative to elogind, which is quite big itself.
SUVs are being sold like candy, and they are heavier than your regular sedan. This is especially so in the US, where it's not just SUVs but also trucks, and they are both bigger and heavier than in Europe. Being heavy is not exclusive to EVs.
I'm on that project myself, we are also dealing with simulating circuits, buuut I feel like we could improve our approach somewhat lol. For example, we had a bunch of trouble trying to speed things up, I immediately thought of trying to make most things run on a GPU but we quickly found out that our circuits just have too few qubits to be parallelized decently.
More importantly, I am not very optimistic (to say the least) about the short- to medium-term real-world applications of quantum circuits we are looking into (we do time series classification, which is quite removed from other domains which work better on these computers), and I got the same feeling from the literature. Should I be feeling differently?
I find it incredibly funny that this appears a couple of days after I finished my master's thesis. I realized I made a bit of a blunder a week before handing it in: I thought the Hilbert transform gets rid of the aliasing when a real signal's bandwidth reaches below the frequency origin and into negative frequencies, which it doesn't. The bandwidth "folding" is still there, but the negative frequency components are gone (which I did know). Since I got rid of the negative frequencies because of the symmetry of real signals about the frequency origin, there wasn't any point in doing the Hilbert transform. Thankfully I checked all the preprocessing steps I carried out in the thesis, and weeded out this unnecessary step.
In my defense, please do bear in mind that I have not had a thorough education in signal processing, physics degrees don't usually have courses like this. I know Hilbert spaces much more well than I do the Hilbert transform.
I just so happen to be working on that at my company, but sadly we do not seem to be hiring more for this position, and I don't think you'd like how low salaries are in southern Europe.
Radar signals are modulated in very specific ways, which are visible right away even with plenty of noise on a spectrogram. Classifying the modulation of radar signals is something common in military contexts, since it allows you to listen for emissions and be able to tell if it's an enemy or an ally. I bet it has more uses than that, but it's the first one I could think of.
We do not know if and I bet plenty of people do not believe such negative energy really exists.
Besides, all warp bubble metrics so far are inertial; meaning the ship is unable to accelerate, so they're a gimmick nowadays. I have hopes we find something workable though.
Quantum algorithms analogous to convolutional neural networks are O(log N) instead of O(N), so I'd wager we really do need something like that, or at least some other sort of analog hardware, given how complicated it is to do calculations on a QC.
Of course you can. Otherwise we wouldn't use fiber optics everywhere. You only need a teeny tiny amount of power transmitted to carry a distinguishable signal after all.
>I love that there are people far more educated than me who are also paid to explore these topics.
Sadly, there's some of us who would explore these topics, yet the pay and some of the conditions surrounding this profession are less than ideal, so to speak.
Also, you might want to study the math around the strong interaction and gravity before making any speculation like that. What is explained to you in words might not be accurate.
I'm currently in a master's degree in data science and I haven't really felt like what you said was happening; we were taught just the right amount of theory (how things like backpropagation, PCA, statistics, optimization work) but more advanced techniques were described only briefly, just so we could understand how to use them in our assignments, which makes sense given the little time there is. There was much emphasis on making use of a wide variety of techniques in practical scenarios.
By reviewing the source code of the model, reviewing the training data, and reviewing weight initialization, but the latter should be specified in the source code. Also making it abundantly clear that the libraries used to make the model were not tampered with, maybe hashing their files or doing some reproducible builds wizardry...
Edit: Now that I think about it, can't data poisoning happen when predicting, rather than just happening in the training phase? In that case, it's going to be complicated to work around that.
It sure looks like such models are going to have to undergo the same sort of scrutiny regular software does nowadays. No more closed-off and rationed access to the near-bleeding-edge.
Get rid of page limits on papers or at least encourage researchers to leave links to further resources on their work. More than once, I have had to complete derivations shown in papers which had some missing steps. So, if access to full, well-explained derivations was widespread, researchers wouldn't have to spend as much time trying to understand other people's work. Sadly this is not seen as much of a problem as it truly is.