tnecniv

26 ms·

331.
by tnecniv 3y ago
This is an aside but the MLB has been actively modifying rules the past few years. Some of them have been quite good, but others seem to contradict each other in purpose and leave me wondering what they actually want. For example, they have
332.
by tnecniv 3y ago
You could define the Big Bang to be at any value of t, i.e., pick t0 > 0. Such a shift changes nothing and won’t fix any singularities in the dynamics.
333.
by tnecniv 3y ago
Do you have any recommended reading on critical phenomena? I’ve become quite interested in statistical mechanics lately, but my background is engineering / applied math. Thus, I learned information theory first and am interested in how
334.
by tnecniv 3y ago
Not the author but it seems like a thermal transition, given that they’re analyzing how the potential well evolves over time. Another interesting thought your second paragraph gave me is if you could do any RG stuff, which has had a lot of
335.
by tnecniv 3y ago
Interesting paper. I’ll give it a full read later. I can’t say I’m surprised because diffusion models are really just a thermal process transitioning between equilibria.
336.
by tnecniv 3y ago
PINNs are neat but by my understanding they’re conceptually not too different from traditional numerical solution methods. The solution to your PDE is some complicated function that is not available in closed-form, so you require a numerica
337.
by tnecniv 3y ago
They’re fun and comfortable.
338.
by tnecniv 3y ago
There’s a Philip K Dick novel where the decay of Earth has led to the colonization of the solar system, but most people on the colonies are barely surviving in hostile living conditions. Their main form of entertainment is to do drugs that
339.
by tnecniv 3y ago
Yeah it’s supply and demand. There’s a lot more people that would love to be concert pianists than there are concert pianist jobs. I feel like these conversations often ignore the middle class of musicians that make a decent living gigging,
340.
by tnecniv 3y ago
This is about 12 orders of magnitude off what you get using the mass-energy-information equivalence. The MEI is theoretically what you arrive at by combining E = mc^2 with the the fact that the entropy of a system at a fixed temperature can
341.
by tnecniv 3y ago
Science is self-correcting. It just might ruin a number of grad students lives along the way.
342.
by tnecniv 3y ago
There’s no reason you can’t do both. Summarizing Tufte: you should always show the data if possible, but you can both show the data and analysis of the data. Don’t throw the baby out with the bath water.
343.
by tnecniv 3y ago
Familiarity for the researcher breeds contempt. If you succeed once and don’t get caught, it’s easier to do it the second time. It doesn’t require a culture of dishonestly for bad actors to get comfortable breaking rules.
344.
by tnecniv 3y ago
Well we don’t know how many people falsify data because, if it goes undiscovered, we don’t know it happened. It could easily be that the rate at which data is falsified is quite low and we are good at catching it, or it is high and we are b
345.
by tnecniv 3y ago
I forget the guys name (he’s German iirc), but a very impactful one was at Bell Labs in the 00’s. STEM PhD programs now all have a research ethics component because of him. While you can’t teach someone to be ethical, it at least hammers ho
346.
by tnecniv 3y ago
While fields are small, academia is a big place. Most academics won’t falsify data — from a pragmatic perspective, that’s their career on the line, and from an emotional perspective, their academic work is tied to their self-image. It’s als
347.
by tnecniv 3y ago
The problem with a lot of these systems is that copyrighting happens at the time of creation whether you register or not. If you don’t register, it may be harder to prove someone is violating your copyright in court, but it is still the IP
348.
by tnecniv 3y ago
This is actually a big issue with academic research related to bounded rationality. Although you could model it mathematically in another way, by far the most common is to use the rate-distortion approach. Rate distortion theory basically b
349.
by tnecniv 3y ago
What you’re basically describing is bounded rationality, which has been widely studied in behavioral economics, psychology, and engineering applications (Simon and Gigerenzer are two big names to google). A common framework for formalizing
350.
by tnecniv 3y ago
Well in Suits they explicitly only hire Harvard Law grads, which is a big plot point and they note multiple times that this is a dumb policy they have for traditional and marketing reasons
351.
by tnecniv 3y ago
That’s a shame. As scummy as he is those videos were excellent. Just how fast he was able to get data into excel and do things with it seemed like magic.
352.
by tnecniv 3y ago
I’ve periodically looked into it the differences and mostly it seems to come down to priorities. Statistics prioritizes mathematical rigor all the way through the process from collection to analysis. ML / Data science has more algorith
353.
by tnecniv 3y ago
Roller Coaster Tycoon was written in assembly despite it being the 90s. That’s a lot crazier than writing a platformer in C. Many, if not most, are C++ with scripting. If you’re working on something that doesn’t need scripting, pure C is to
354.
by tnecniv 3y ago
An interesting idea. My first thought was to say “just use a differentially private learning algorithm,” but DP requires a notion of closeness, and that’s not trivial for, e.g., image data. Images close in an abstract sense, like Euclidean
355.
by tnecniv 3y ago
Like dynamic typing and other things, declaring things in the middle of your code block seemed like a great idea to young me, but as I’ve gotten older, things like organizing variable definitions, strict use of types (or the kind of sucky t
356.
by tnecniv 3y ago
PhD students are almost always funded including an annual stipend. The university charges tuition but almost none of them ever see that bill (their advisor or external fellowship does). Undergrads pay for their education mostly
357.
by tnecniv 3y ago
I am so excited to play this tomorrow
358.
by tnecniv 3y ago
I can’t wait for the Eugenics Wars, myself
359.
by tnecniv 3y ago
What’s your solution for hosting the archive of the data if you want to share it? That’s a weakness in my workflow.
360.
by tnecniv 3y ago
Yeah if your analysis to takes hours to run, you should really split up the number crunching code and result analysis / visualization. Not only does it make version control of the code easier, you can save the output in an organized la

More ›