cameronperot

One could, but automatic differentiation is much more efficient than numerical differentiation, thus for high performance applications it is preferable to use automatic differentiation.

Just donated as well :)

I absolutely loved my WRT54G series router. I had one years ago that had its input ethernet port fried during a storm. Luckily I was running DD-WRT on it and was able to reconfigure one of the output ethernet ports as the new input so the router lived on.

Looks like a decent basic summary. Section G.2.6 should probably be called "Multiple Dispatch" rather than "Function Overloading" though. Defining multiple methods of the same function in Julia is different to function overloading in a language like C++ [1].

[1] https://www.youtube.com/watch?v=kc9HwsxE1OY&t=392s

I have been using Signal since before it was even called Signal (TextSecure and RedPhone), and I must say it has really refined itself over the years. It's pretty easy to set up these days, and I have more non tech oriented friends switching over seamlessly. I use Signal's voice and video calling features as well without any issues.

Yes, you can enable the "default encryption module" in the apps page like any other Nextcloud instance, as well as other apps, e.g., 2FA.

I would suggest Hetzner's storage share (their managed Nextcloud hosting) [1]. Their pricing starts at €2.90/mo for 100GB of storage, which is significantly cheaper than what is mentioned in the post ($5/mo for 25GB).

[1] https://www.hetzner.com/storage/storage-share

I've used Firejail for years and can definitely recommend it! It has a lot of nifty features, e.g. network namespacing via the --net argument.

For anyone interested in additional resources, MIT has some courses taught by Peter Shor and Isaac Chuang on quantum information science [1-6].

[1] https://www.edx.org/course/quantum-information-science-i-par...

[2] https://www.edx.org/course/quantum-information-science-i-par...

[3] https://www.edx.org/course/quantum-information-science-i-par...

[4] https://www.edx.org/course/quantum-information-science-ii-qu...

[5] https://www.edx.org/course/quantum-information-science-ii-ef...

[6] https://www.edx.org/course/quantum-information-science-ii-ad...

PrivacyTools usually has an up to date list [1] of good extensions to use. It might also be worth going through their about:config tweaks section [2].

[1] https://privacytools.io/browsers/#addons

[2] https://privacytools.io/browsers/#about_config

[4] is the wrong link, but it's too late to edit. Correction:

[4] https://ocw.mit.edu/courses/physics/8-06-quantum-physics-iii...

It was merely a suggestion, I never said it was necessary. However, in my experience, having knowledge in quantum physics has made it easier for me to read literature related to quantum computing, hence why I suggested it.

I would suggest starting from the basics of quantum physics if you don't already have a solid foundation. MIT has several excellent courses on the subject [1-4]. The first two are the same course, but taught by different professors with slightly different curriculum.

[1] https://ocw.mit.edu/courses/physics/8-04-quantum-physics-i-s...

[2] https://ocw.mit.edu/courses/physics/8-04-quantum-physics-i-s...

[3] https://ocw.mit.edu/courses/physics/8-05-quantum-physics-ii-...

[4] https://ocw.mit.edu/courses/physics/8-06-quantum-physics-iii...

I came across that paper just recently, it was a very good read!

I'm not working on any specific problem yet, but for my master's thesis I'm hoping to do something related to the use of neural networks in numerical solutions to differential equations. Along the lines of this sort of stuff [2].

[2] https://arxiv.org/pdf/2001.04385.pdf

I'm studying in the intersection of physics and data science, and I think there's a number of places where physics can benefit from ML. From my current point of view though, most of these applications lie more on the experimental/computational sides of physics rather than the theoretical side. One of the current use cases is using ML to aid in the processing and analysis of data obtained from experiments.

I would like to see more truly innovative work done on the theoretical side, but I don't think we'll see "AI" bridge the gap between QFT and GR any time soon. I think in order for something like that to happen we need a new approach, as the current approach of throwing deep learning models at it doesn't feel like the right answer.

On a more general note, the SciML organization [1] has been quite successful in helping incorporating more ML into science.

[1] https://sciml.ai/

I've sold old computer hardware on both eBay and hardforum [1]. The nice thing about hardforum is a lot of buyers have heatware [2] which makes it easy to judge their legitimacy. I'm sure there's other forums out there that have similar buy/sell sections you can use. To increase your legitimacy it's best to maybe explain a bit about the hardware (e.g. how hard was it run) and have a picture with something showing ownership (e.g. the hardware with a piece of paper where your username is written).

[1] https://hardforum.com/

[2] https://www.heatware.com/

I do a lot of scientific computing, mostly in Julia, and having a REPL at my disposal is extremely useful. There's a great VIM plugin [1] that enables you to send code from the editor to a terminal to be executed, e.g. a tmux session which has a REPL open. It can be coupled with other nice plugins such as vim-ipython-cell [2] and vim-julia-cell [3].

[1] https://github.com/jpalardy/vim-slime

[2] https://github.com/hanschen/vim-ipython-cell

[3] https://github.com/mroavi/vim-julia-cell

That's a tough one. On one hand it should be easy because who wants to perform more clicks to get the same end result. On the other hand it shouldn't because it attracts people without proper knowledge who could possibly end up making a life altering mistake. I'm not really sure what a good compromise is here.

The only thing Robinhood did good was pushing other brokerages to get rid of commissions.

I would hope that most people left their platform after they had those major outages during volatile trading days a few months ago. Any place that has outages lasting that long without properly explaining what happened shouldn't be licensed to deal in securities.

Now they're just preying on people that aren't knowledgeable in the financial markets by showing them how "easy" it is to trade. I have studied and worked in finance, and I'll be the first person to tell you, trading isn't the least bit "easy" and shouldn't be portrayed as such.

I had to turn off uBlock Origin to get it to work.

After living in Germany (originally from the big city in Texas), I've come to realize how valuable it is for a city to be walkable.

I really like being able to walk out of my door and reach anywhere I need to on foot, especially when there's a lot of beautiful nature around. I usually walk in the evenings for at least an hour (outside of any walking I might do in the day to go to the store or such) as a way to relax.

In theory, yes, you would want just enough thermal paste to fill any gaps between the heatsink and die as the thermal conductivity of the heatsink (e.g. copper) is much higher than that of the thermal paste. Unfortunately, I haven't been able to find much on the internet beyond forum posts and anecdotal evidence regarding the pump-out effect. It would be interesting to see a more scientific study of this.

Yes it does.

Good to know. If the PK-3 ever fails me I'll give the Kryonaut a shot.

Thinkpads with a dedicated GPU come with a dual heat pipe heatsink, whereas ones without the dedicated GPU don't. So, you can take the heatsink for the former and use it on the latter, given that you make sure to insulate the part of the heatsink that is supposed to cool the GPU (this can easily be done by using a quality electrical tape like super 88 on the bottom side). You can buy the part aftermarket or see if Lenovo is willing to replace yours. The part number for the heatsink I put on my T490 is 01YU186.

Interesting, I had heard somewhat of a different take from Der8auer (who is associated with Thermal Grizzly). In [2] he explains that Kryonaut is a fairly wet thermal compound and that it does indeed suffer from the pump-out effect, and that in direct die applications like GPUs they recommend replacing it once per year.

[2] https://youtu.be/CCqxE-5Ct3w?t=505

I replaced the single pipe heatsink on my T490 for one with dual heat pipes as the single couldn't sufficiently cool the CPU (was hitting 97C out of the box and after several re-pastes). So I set out on a journey to find a good thermal paste to use. However, finding a good thermal paste for laptops (more specifically on die usage) isn't as straightforward as one would think. A lot of the high end thermal pastes work best for desktop use, i.e. between a heatsink and an IHS, whereas on laptops you don't have an IHS so it's between the heatsink and the die. These high end thermal pastes usually all succumb to the same problem called the pump-out effect [1], where after a number of thermal cycles the performance of the paste begins to fall and it needs to be re-pasted.

I ended up going with Prolimatech PK-3 as it's a fairly viscous thermal paste, and it has held up well for the past six months. Although, I have noticed slightly higher max temperatures than when I originally applied it, but nothing alarming.

[1] https://www.overclock.net/forum/28099300-post10.html

Gilbert Strang covers a this well in his "18.065 - Matrix Methods in Data Analysis, Signal Processing, and Machine Learning" course [1]. I can highly recommend the lecture series to anyone interested in such topics.

[1] https://www.youtube.com/watch?v=rYz83XPxiZo&list=PLUl4u3cNGP...

A little late in the game, but they're a brand I would hold in higher regard than 99% of the other providers out there. I believe that a lot of people misunderstand what exactly a VPN is and what scenarios it offers benefits of use in. I personally host my own VPN on a lowendspirit server [1] for when I'm on an untrusted WiFi network or I need to have an IP in the US (it comes in handy as a US citizen living abroad). I also use a VPN sometimes when I have a dev server (hosted on the server itself) that I'm developing/testing on since being on the same network as the server makes things easier, e.g. having a container with an API bound to the VPN network so that I can access it easily and without it being public facing.

Of course there's also the shady side of VPN use. If you're doing that it might be beneficial to use the VPN within a VM with strict firewall rules, i.e. only allow incoming/outgoing to/from the VPN. Doing so allows you to only send the traffic you want to over the VPN, thus reducing your exposure to any nefarious data collection that the provider might be doing.

[1] https://lowendspirit.com/