They should have asked a particle physicist...
In other words, this is a situation where a few events must be found inside a very large data set, and the signal looks remarkably similar to the noise. That makes it quite difficult to apply machine learning, let alone train the algorithm in the first place.
Uh, what? The algorithms are trained on simulations, where you have as many events as the available computing time allows.
Just that tiny portion of the data must be refined from a "may contain traces of Higgs" state to a definitive "contains Higgs" or "Higgs free."
That sentence switches from an event-by-event basis (first part) to the overall analysis statement (second part), in the way it is phrased it is very misleading.
two gluons collide, and via a love triangle of two virtual top quarks and a virtual anti-top quark
It doesn't make any sense to say "two virtual top quarks and a virtual anti-top quark".
The detector only registers the following: the energy of the photons and their angle with respect to the beam. From that, you can figure out how much momentum, in the direction perpendicular to the beam, the photons have (called transverse momentum).
The azimuthal angle (##\varphi##) is measured as well, otherwise you couldn't reconstruct the mass.
Oh, and the detector records much more to distinguish between photons and other objects or to find the origin of the photons.
So far so good. But there are a bunch of non-quantum machine learning algorithms that should be able to do the same.
What do you mean by "should be"? ATLAS and CMS have used these algorithms for years.
The researchers chose several of them and set them loose on the data.
That's what ATLAS and CMS did already.
or algorithms trained on around 200 collisions, the quantum algorithm significantly outperforms the classical algorithms.
Fine... but no one will use a training dataset of just 200 collisions.
On the other hand, the quantum algorithm is significantly worse than the classical algorithms after training on large data sets. This, however, is probably a product of the performance of the underlying hardware rather than the actual algorithm.
In other words, using the quantum computer made it worse.It is worth noting that the Higgs can be found in the two photon decay mode without any machine learning. Machine learning just improves the sensitivity. The Higgs discovery was announced early July 2012, without machine learning it might have been late July 2012 or August 2012, probably not later. Not as important as the article might suggest. There are other studies where it is more important.
The
original study looks fine, just the report about it is not good.