Skip to main content

Machine learning to help Higgs experiment

The University of California, Irvine, (UCI) has published findings in Nature Communications that suggest that algorithmic deep learning with digital images could greatly increase the odds of finding the Higgs boson – the recently confirmed particle that could explain fundamental principles of the universe such as why some particles have mass.

In computer experiments using carefully structured simulated data, the UCI researchers' methods resulted in an 8 per cent increase in the detection of these particles.

Machine learning, the umbrella term that encompasses deep learning, is a branch of artificial intelligence whereby computers are taught to recognise patterns in data. Finding Higgs boson particles requires sorting out relevant data from huge amounts of background noise.

Pierre Baldi, Chancellor's Professor of computer science at UCI, commented: ‘It's very difficult to write from scratch a program that can recognise elephants in images – or Higgs bosons in collider data. But we can provide to the computer many examples of images with and without elephants, or accelerator data with and without Higgs bosons, and let the computer learn automatically from these examples.’

In May, Cern and other organisations launched the Higgs boson machine learning challenge, a competition to develop machine-learning techniques to improve the analysis of Higgs data.

The competition is run through Kaggle and applicants are asked to design a system that can separate simulated images into either ‘tau tau decay of a Higgs boson’, or ‘background’. The competition closes on the 15 September, with the top three scores receiving cash prizes.

Baldi, along with computer science PhD student Peter Sadowski and associate professor of physics and astronomy Daniel Whiteson, found quicker, more efficient ways to analyse data obtained from particle accelerators or colliders to better detect rare particles.

Currently, physicists devise by hand mathematical formulas that they apply to the data to derive the features they're looking for, which are then fed to machine learning programs. By employing recent advances in deep learning, in which computers learn automatically at multiple processing levels, the UCI team eliminated the need for the time-consuming manual creation of those formulas.

‘These new, smarter deep learning networks have shown themselves to be better at finding hints of new particles than past machine learning methods – and than physicists with years of experience,’ Whiteson said. ‘They don't need any help from human insight, achieving a level of automatic learning which has been a long-standing goal in high-energy physics.’

Further information:


Read more about:


Media Partners