DeepSea Technologies is to act as the maritime industry representative in a new initiative called the Shifts Project, which will bring together a variety of institutions and researchers in the field of AI across different sectors to study and solve problems associated with distributional shift.
Distributional shift refers to the changes found between the data distribution used in an algorithm’s training dataset for AI and machine learning applications, and the dataset the AI encounters when actually deployed in real life scenarios.
Maritime has been chosen as one of two case study challenges for the project, with the other being distributional shift in relation to the treatment of the chronic medical condition Multiple Sclerosis.
Alongside DeepSea, the Shifts Project includes the universities of Cambridge, Basel, Lausanne, and HES-SO Valais, bringing together core machine learning (ML) researchers studying distributional shift with applied ML researchers, who work on tasks affected by distributional shift in the real world.
One example of distributional shift that DeepSea has highlighted in maritime is changes in vessel performance as a result of hull fouling, which can lead to various operational inefficiencies. Understanding how the entire vessel’s performance data shifts over time is critical to accurately modelling vessels, and consequently using those models to assist in decarbonisation and minimising fuel waste.
“This is an internationally important research field for all of AI, and managing distributional shift is a topic that we have been focused on since our inception. It’s a critical prerequisite to using the technology to generate real impact for companies both now and in the future,” said Dr Antonis Nikitakis, DeepSea’s Head Research Scientist.
“This project follows on from our research that was announced at HullPic in May of this year, which focused on creating rigorous standards for AI within the industry. This was in response to the growing need of consumers to fully understand what they’re getting when they consider adding AI as an optimisation tool.”