mattheww

Did my thesis research at Brookhaven National Lab, home of the Relativistic Heavy Ion Collider (RHIC), which is the predecessor of the heavy ion program at the LHC.

While there, one of the more senior scientists relayed an exchange from an ongoing review of the program. At the time, RHIC was colliding gold in the heavy ion program.

One of the reviewers asked if RHIC could save money by switching to a cheaper element, like lead. None of the RHIC representatives knew what to say. I don't remember the exact numbers, but RHIC used something like < 1 milligram of gold over the lifetime of the program.

You're right, the thing limiting the value is not demand - there are more than enough faculty and high-level staff that want to live there.

The thing limiting the value is the other restrictions that Stanford imposes on these houses - namely they essentially control the price the houses sell for because they all have to be financed through a Stanford-controlled lending program.

FYI, if you make a realistic calculation (including tax and multiple winners), the EV of a lottery ticket is still negative.

Depending on your tax bracket, the non-jackpot prizes have about $0.19 of EV.

In a state where you don't pay tax on lottery winnings, like CA, you cash option post-tax is currently estimated at $408,403,045.

My estimate of the multiple winner correction to the EV is 0.8, based on numbers of winners for prizes >$500M.

Your jackpot EV is therefore $1.08 and the total winnings EV is therefore $1.27, meaning the overall EV is -$0.73.

For there to be positive EV, that delta needs to all come from the jackpot, so the headline jackpot number would need to go from $1.1B to $1.8B to get it there.

There are four questions at the end of the post.

For sure, the second one is answered - it is possible to parallelize GNNs to the billion-scale, while still using message passing. It requires rethinking how message passing is implemented, modifying objective functions that work in parallel, and changing ML infrastructure. You're not going to get to large graphs with generic distributed Tensorflow.

I don't know if the third question is fully answered, but there are many approaches to preserving locality, either by changing architectures or changing objective functions.

Also, errata: PinSage was developed for Pinterest, not Etsy (hence, not EtsySage).

It's unbelievable that nobody has done this well - I and most people I know are tracking their own job searches on a spreadsheet.

Pretty obvious value-add for job searchers. Not to mention that having access to this data would enable tons of other product features. Shows that most services/sites don't care that much about the applicant experience.

If you know a way to find clusters of intersections of hyperplanes, I'm pretty sure you can get a highly acclaimed paper out of it, but that is not what a Hough transform is. The Hough transform is an approximate solution to that problem which works by sampling the hyperplanes and then polling them. There's no way to perform a Hough transform without having many more transformed points than input points, and the more precision you need, the more points you need.

To be fair, 50 PU is above design peak luminosity, much less mean. And I'm sure I've seen plots from both ATLAS and CMS at the end of LS1 that show improvements in the processing time at 100 PU by factors of roughly 10.

Hough transform is basically useless for precision tracking in a high multiplicity environment. Tracks have 5 degrees of freedom, so the memory costs make it infeasible. I think ATLAS uses it in the trigger, where you don't need to actually reconstruct all tracks, just find out if there are a couple passing certain criteria.

Depends exactly how you limit the problem, but going from detector hits to abstract arcs takes a few seconds.

The main thing of interest here is whether a system that knows about physics (traditional approaches) can be beaten by a system that has no a-priori knowledge of physics (OOB DL) or if someone will find a way to integrate physics knowledge into a DL approach.

RE 2: What I mean is that "N-sigma" and local p-value are being used interchangeably, which you can see from this plot. There's no "number ... calculated without including info about any theory/evidence regarding the Higgs boson".

RE 3: The model used to calculate the background are unphysical in the sense that they set the Higgs production cross section to zero without changing anything else. There's no physical Higgsless theory.

RE 4: I was agreeing with you and adding some facts.

That number is calculated without including info about any theory/evidence regarding the Higg's [sic] boson.

That's not true. You can see plots in the 2011 and 2012 papers that give these calculations as a function of mass. It's basically impossible to make a mass-independent calculation. (Not withstanding the fact that there's no Higgsless theory to do calculations for the background).

As an example, I'd imagine detecting the Higg's [sic] boson at 1 eV energy levels is theoretically predicted to be even less likely than billions/trillions to one odds, therefore detector noise would be a more likely explanation for such results, despite the low p-value.

You're right, the p-value is much lower. Previous experiments have long since excluded such a low Higgs mass. Also, if the Higgs mass were so low, we probably wouldn't exist.

So there's no obvious practical application (right now).

Also, the proton is not 4% smaller. Protons are obviously whatever size they are.

The discrepancy comes from the fact there are two techniques to measure the proton size. Both experiments do their thing and then there's a way to interpret the results that would tell you the size of the proton (look up proton form factors).

However, when you do the interpretations, which depend on some theoretical calculations, you get different results. The general thinking around this result, because nobody has found any issue with the experimental results, is that there are some additional interactions that are stronger than expected that need to be accounted for (there are some unknown quantities that allow this).

One of the interactions would only affect the muonic hydrogen measurement - basically there are some different interactions between muons and protons than between electrons and protons because of the muon's mass and those might be different than originally thought.

The other is a type of interaction that could affect both normal and muonic hydrogen. This new measurement shows that the interactions that affect both has to play an important role in understanding this discrepancy. There are other measurements trying to measure this effect independently (not using hydrogen at all).

As you can imagine, the long-lived particles can have "any" lifetime. Since the lifetime is unknown, it makes sense to probe many lifetimes. The experiment being suggested by Lubatti and his collaborators probes lifetimes on the order of 10^5 - 10^7 m. There aren't strong constraints on this range, which can appear as missing energy in CMS or ATLAS (or not). CMS, ATLAS, and LHCb are giving constraints at 10^-6 to 10^2 m. Interestingly, the different experiments are probing mostly different ranges.

To correct something in the grandparent (?) post, the LHC experiments do not know the incoming energy. Even though the beam energy is 6.5 TeV on 6.5 TeV, the collisions are only a fraction of that. We only know that the energy has to balance in the perpendicular direction. However, if a collision produces two invisible particles that balance each other, it would appear that there is no missing energy. In many of models, pair production of new particles is preferred, so if the particles are long-lived, they can be hard to find.

The argument in favor of an electron-positron collider is that there is much less "junk" in the detector from the hadron collision. So the trade-off is that you produce a lot fewer Higgs bosons, but, in principle, you are able to measure each Higgs much more precisely. Having a machine at 250 GeV puts us at the sweet spot for producing the Higgs with mass 125 GeV.

There is some discussion in the community about whether it truly is going to advance the field to build such a machine. It's not super clear whether the various proposed 250 GeV machines will improve on what will be done by HL-LHC. From the Chinese point of view, though, it absolutely is the right decision to build this machine on the way to a 100 TeV hadron collider, since they desperately need to build up some local expertise in constructing/operating a large collider.

Experiments like IceCube answer different questions than accelerator based experiments. You couldn't, for example, measure the Higgs branching ratio to bottom quarks at IceCube. On the other hand, you can't measure high-energy neutrinos at the LHC. Both types of experiments have their place.

This is not a very straight-forward question to answer. The short answer is that both CMS and ATLAS detected a couple thousand Higgs events.

The long answer depends on your definition of detected. As you probably know, the Higgs decays into many different combinations of particles. In some of these cases, we have very little chance of determining whether there was a Higgs in the event. In fact, only in H(iggs) -> ZZ -> 4l(eptons), do we stand a reasonably good chance of saying whether a specific event contains the Higgs. In this channel, CMS and ATLAS both observed ~20 events.

The next two most sensitive channels are H->WW->2l+2nu (neutrino) and H->2photon. In the most Higgs-enriched regions that have been constructed, the contribution from Higgs is ~10% in H->WW->2l+2nu and ~20% in H->2photon. So in these channels, we have very little chance to say whether a specific event contained a Higgs, but if we look at all the events, we see features that wouldn't be present if there weren't any Higgs bosons. The contributions from Higgs in these channels are a few hundred events in CMS and ATLAS.

Then there are even less sensitive channels, which means we have even less chance to say whether a specific event contains a Higgs. CMS and ATLAS probably detected a few hundred to a thousand or so events each in these channels.

I am a scientist, and I have seen a lot of terrible code. Most scientists have no formal training in computer science or coding. Many advisors don't place much value in having their grad students take such classes, though even a short language-specific introduction class would vastly improve their students' productivity.

I recently undertook a complete rewrite of our group's analysis software that was written by our previous postdoc. It was ~30k lines of code in 2 files (one header, one source file), with pretty much every bad coding practice you can image. It was so complicated that that postdoc was essentially the only one who could make changes and add features.

The rewritten framework is only ~6k lines of code to replicate the exact same functionality. It's easy enough to use that just by following some examples, the grad students have been able to do implement studies in a couple days that took weeks in the old framework. The holy grail is for it to be easy enough for the faculty to use, but that will probably take a dedicated tutorial.

My point is that following "best practices" may be overkill, but taking a thoughtful approach to the design of the software can vastly improve your productivity in the long run. Posts like the OP help scientists who write bad code defend poor practices. Any scientist worth his salt should support following good practices because it will always lead to better science.

It's based on the fact that particle energies follow a continuous distribution.

See, for example: http://www2.astro.psu.edu/users/nnp/cr.html.

Close to nothing would happen. The relevant quantity to understand is the stopping power (usually referred to by us physicists as dE/dx). Stopping power drops significantly as energy of the charged particle increases. So for very large energies, very little of that energy is deposited over a given distance (in an absolute sense).

This principle is how some radiation therapies work to treat cancer. Because the stopping power curve for protons is well understood, a proton beam can be tuned to deposit all of its energy is a fairly small area.

Aside/Rant: this question clearly can be answered by someone with expertise in the field. So why did people feel the need to speculate about it instead of just waiting for someone who knows how to answer it?

Actually, it does follow. If there are particles of the energy described in the original link, there are many more particles of lower energies also hitting the Earth's atmosphere, including energies accessible at the LHC. And none of these resulted in the Earth's destruction.

The same logic applies to the argument that any black-holes or otherwise dangerous particles would simply zip by the Earth (stated in a sibling comment to yours). Particles of slightly lower momenta come in higher numbers, so these would produce the dangerous particle, but not retain enough momentum to escape. Since the Earth has not been destroyed, there are no particles produced at low enough energies for us to destroy ourselves.

These off-base criticisms reflect the fact that commenter did not read the whole paper and likely did not even read the introduction, which ends on page 3, where he found "serious" problems.

For example, the paper nowhere conflates consciousness and memory. Instead, the paper repeats the suggestion that a requirement of consciousness may be the ability to process a substantial amount of information. This idea is related to the idea of memory, but it's not really the same.

The paper is not about describing the biological experience of consciousness at all. This paper asks the question: "Is there some way to understand from the Hamiltonian and density matrix of our universe that it should contain consciousness?"

Tegmark then proposes some criteria that are probably necessary for consciousness to arise and then presents some metrics for the various criteria and calculates those metrics for various conditions. The paper is really describing a framework for how consciousness can be considered in the context of a physical representation. The calculations should be relatively straight-forward to follow for anyone with a decent memory of their linear algebra class.

To put it another way: If you had a Hamiltonian and density matrix that described a universe that you thought contained consciousness, what kind of things would you calculate for that Hamiltonian and density matrix to try and find out if it did or didn't? This paper suggests some ways to think about this question.

So I don't understand why this comment spends so long talking about "observing consciousness."

Why do you want to limit by location when the the data is already separated by cause? All of the Tesla fires were caused by collision.

If you compare (Fires caused by collision)/Vehicles on the road, then I agree Tesla does not compare favorably. In principle, you should compare Fires/collision in case Tesla drivers have an abnormally high collision rate. And statistically speaking, it's pretty hard to draw a conclusion when your sample size is 3 fires.

But the fact is there are still no fires attributed to electrical or mechanical problems.

And the points on both sides about injuries/deaths from fires are pretty much red herrings since the numbers are already extremely low. There are only 2 deaths per 100 fires caused by collisions.

I will explain why lisper's explanation of light speed is wrong. The fundamental problem is that it violates the premise of relativity and is a classic example of appealing to an absolute frame.

Velocities are only measured relative to other objects. The explanation given by lisper requires an absolute reference frame. Otherwise, you have no way to figure out how much you're actually moving through space, and thus, through time because you can never figure out your "true" speed.

Relativity tells me that no matter what speed I travel at, I will experience time at the same rate. From my frame of reference, I will perceive things to happen in other reference frames at different rates. When we calculate time dilation and length contraction, we're determining that if something took time "t_a" in Frame A, it will appear to have taken time "t_b" to an observer in Frame B.

Furthermore, it's impossible under relativity to have a frame of reference that moves at the speed of light in any other frame.

There are two separate questions here.

The first question, "is the Higgs mechanism how elementary particles get their mass?" is now essentially answered. Whether or not the Higgs boson is a single particle or part of a hierarchy of related bosons is almost irrelevant to the answer of this question.

The second question is "is there new physics at the TeV scale?" If there is, then there are additional bosons related to the Higgs in some models like supersymmetry. However, it is not required that there be additional bosons or that that they be easily measured. Other new particles might be more accessible. On the other hand, there might not be any new physics at LHC energies at all. This question is what will take time to answer.

Unfortunately, the idea that supersymmetry is "dead" has been propagated by journalists covering science who don't know any better and people with an agenda. Generally, it's people who just don't know any better. Unfortunately, you see comments supporting the idea even on HN.

The important thing to understand about supersymmetry (SUSY) is that in the most general case, there are approximately 105 new parameters. That's far too many to probe in a meaningful way, so most models choose between 2 and 5 to vary, and fix the rest. Then a bunch of models are chosen that hopefully cover a wide spread of different behaviors.

The true part is that several models have been excluded, basically as well as the LHC is going to be able to exclude/discover anything in the current energy regime. However, some of these models were just not chosen very well to begin with (but have historical importance) and others were chosen to have maximal signal strength.

So as time goes on, the search for SUSY turns away from "easy" models and looks at more complicated ones. With 105 parameters, there's a lot of parameter space unexplored.

There are a lot of people thinking about what kind of structure the model of dark matter particles could have. And there are a lot of different ways for there to be more than one dark matter particle, and even if the only way those other particles interact with normal matter is through gravity, there are still ways for us to understand things about them.

At the moment, the most popular dark matter theory comes from supersymmetry. In this case, there's only one dark matter particle, and all of the rest of the particles interact with normal matter, in pretty much the normal way, since all of the underlying structure of the model is almost identical.

In flights I've been on with obvious unaccompanied children, someone from the flight crew tells the child to stay in his/her seat until everyone else has gotten off the aircraft.

The fact that you think it's being ruled out makes it testable. I'm not sure how many people would agree with your assessment, however. There's so much SUSY parameter space that's still not accessible that it's pretty early to declare the nail in the coffin. In Witten's case, it probably wouldn't affect the case for his award anyway.

From the AdS/CFT correspondence (Maldacena), it's possible to calculate the ratio of viscosity to entropy in the quark-gluon plasma. This ratio has been measured at heavy ion colliders (well at least one heavy ion collider).

Both Witten and Maldacena have made contributions that are being tested today. I'm not familiar with Sen's or Seiberg's work, but just from the summary, I can tell you that Sen's work is probably in the same boat.