Machine LEarning in Optical NeTwORks

Optical fiber networks is one of the major drivers of our societal progress and a key enabler of the global telecommunication infrastructure.

Optical networks underwent considerable changes over the past decade, as consequence of a continuous growth (exceeding 20% per year) of bandwidth demand.

The current growth sets strong requirements in terms of capacity and costs for the operators, which seek to decrease the cost per transmitted bit.

Several solutions have been proposed, and among them wide-band is more favourable to network operators, compared to more/or novel fibers.

However, wide-band optical system presents new major challenges: optical components must guarantee similar performance over a broad spectrum, network optimization is carried out on a non-flat spectrum and with a much larger number of channels making design, optimization and control a complex problem.

Therefore, application of machine learning (ML) techniques is of the growing importance for high-capacity multi-band (MB) optical systems.

ML is becoming the technique of choice to solve complex nonlinear technical problems, such as, advance component design and management of wide-band networks.

The European Industrial Doctorate MENTOR presents a timely proposal to train 6 ESRs in the interdisciplinary field of high industrial importance: ML applications in multi-band optical communications.

As ML can properly works only when a large amount of real data is available, it is crucial to bring together academic partners and the industry that provide access to the data.

MENTOR consortium offers the strong industrial commitment of four large companies (CORIANT Germany and Portugal, ORANGE Labs and TELECOM ITALIA MOBILE) that significantly contributed in defining the research and training topics to be studied together with the world-leading academic partners in MENTOR.

MENTOR will contribute to the European economy by design of the next generation of high-capacity optical networks.