Musical machine learning tech could be used in industrial and medical applications

Product design and development firm Cambridge Consultants has devised machine learning technology that, it says, can outperform human brainpower when it comes to identifying different musical styles. What’s more, Cambridge Consultants claims, the technology could be equally applicable to detecting faults in an industrial system or rapidly assessing patient health from sensor waveforms.

Live music is complex and unique to each musician – making it often difficult to classify its genre. To put its machine learning system to the test, Cambridge Consultants had a pianist play a variety of music – covering baroque, classical, ragtime and jazz genres – in a live demonstration. Applying complex algorithms, the system then searched for different influences and assessed the likely genre in real time. It overwhelmingly outperformed conventional hand-coded software, painstakingly written by humans.

The breakthrough opens the door to a new generation of multimedia information retrieval systems – providing a sophisticated method of organising and searching music databases.

“Machine learning is at the core of a new wave of artificial intelligence applications limited only by our imagination,” said Monty Barlow, director of machine learning at Cambridge Consultants. “We’re working at the cutting edge of machine learning research and development, developing systems that can apply complex algorithms to big data and ‘learn’ autonomously – without explicit programming.”

Cambridge Consultants’ unique approach involves the creation of what it calls a ‘digital greenhouse’ – an environment where machine learning can flourish. The more strains of models it can grow, the more it can understand where the richest pickings are to be found – beyond the confines of ordinary software development methodologies. The company has an on-site facility with many teraflops of dedicated compute power, where its data scientists and engineers are engaged in long-term machine learning programmes.

“The team is sweating state-of-the-art compute resource to assess new algorithmic hopefuls against long-standing commercial and industrial challenges, such as optimising the deployment of cellular infrastructure or detecting anomalies on a manufacturing line,” added Barlow. “But the digital greenhouse approach doesn’t stop at delivering client projects. It’s an experimental approach that engages the spare time and creative inspiration of our smartest minds, who know they’re working at the frontier of a vital, transformative industry.”