CERN Tests First Artificial Retina For Spotting High Energy Particles

The human retina is remarkably good at pattern recognition so particle physicists are borrowing the design

Pattern recognition is one of the few areas where humans regularly outperform even the most powerful computers. Our extraordinary ability is a result of the way our bodies process visual information. But surprisingly, our brains only do part of the work.

The most basic pattern recognition — edge detection, line detection and the detection of certain shapes — is performed by the complex circuitry of neurons in the retina. Long before an image reaches our brain, the eye has already sorted out many of the significant details.

Physicists and computer scientists would dearly love to have this capability to analyse patterns in experiments. One of the most demanding pattern recognition jobs occurs in particle physics. When subatomic particles are smashed together inside particle accelerators, they create a shower of daughter particles that stream away from the collision. Physicists identify electrically-charged particles by the shape of the tracks they make through detectors that can sense their position, essentially a process of pattern recognition.

The problem is that at the world’s biggest and most powerful particle accelerator, the Large Hadron Collider, the number of collisions is mind blowingly huge. This experiment produces around 40 million collisions per second and each one of them generates huge showers of particles, most of which are entirely uninteresting for physicists.

Every now and again, however, these collisions produce something fascinating, like a Higgs boson. These interesting collisions are a tiny fraction of the total, perhaps as few as one in 10,000. The tricky task that particle physicists face is separating the interesting collisions from the dull ones and doing it in real-time because there is no room to store the data from all of the collisions for later analysis.

For this process, physicists use a triggering mechanism that saves data when it is interesting but discards it when it is not. This process is currently done by superfast electronic systems that measure simple characteristics of the particles involved, such as their energy, and then trigger the data storage mechanism when the energy meets some predefined criteria. This is a crude approach that still ends up collecting much unwanted data.

A far better approach would be to study the shape of the particle tracks in much more detail and trigger data storage only when these tracks are unusual.

That is more or less exactly what the retina does in mammalian vision. It ignores huge volumes of data in the visual field but alerts the brain when interesting patterns appear, such as faces, food and so on.

So an interesting question is whether the same technique could be used to identify interesting particles. Today, a team led by Giovanni Punzi at the Scuola Normale Superiore in Pisa, Italy and a few pals at CERN, Milan and Fermilab, say they have developed the first artificial retina capable of monitoring the firehose of data from the LHC and spotting interesting tracks within it. ”We design a massively parallel pattern-recognition algorithm, inspired by studies of the processing of visual images by the brain as it happens in nature,” they say.

Of course, their artificial retina is a somewhat different to a human eye. The data this retina receives comes from the detectors that surround the collisions. This is like the “eyeball” of the experiment. And the retina itself is a type of microprocessor known as a field programmable gate array, which is designed to mimic the pattern recognition properties of the human retina.

Punzi and co have tested their artificial retina using data that is designed to simulate the data rates that the LHC will produce from 2020 onwards (it is currently being upgraded). And the performance of the artificial retina is remarkable. Punzi and co say it is capable of reconstructing particle tracks at the required rate of hundreds of millions of collisions per second. “This is 400 times faster than any existing or foreseen device used in high-energy physics,” they say.

That is an impressive result. It raises the prospect of a new generation of artificial retinas that will spend their lives peering at the debris created in some of the most powerful collisions in the universe that recreate conditions that existed in the instants after the Big Bang.

However, significantly more work will need to be done before these devices can be used in real experiments. The design of the upgraded LHC was settled long ago so there is no chance of an artificial retina working there in the near future.

However, it may be that artificial retinas are key part of the toolbox that particle physicists use to design the high-energy particle accelerators of the future.

Ref: arxiv.org/abs/1409.1565 : The Artificial Retina Processor For Track Reconstruction At The LHC Crossing Rate

Follow the Physics arXiv Blog on Twitter at @arxivblog, on Facebook and by hitting the Follow button below