Design of new materials and improving of processes, applying novel machine learning methods to sparse and noisy data sets

A UK startup has launched a unique Artificial Intelligence (AI) toolset that can train deep neural networks from sparse and noisy real-world experimental data. Most machine learning methods can only predict or optimise properties for which they have a critical mass of fully-populated training data. Key capabilities that the new methods enable are: Enrich, validate, and understand existing data, maximising return on investment; Guide experimental programs - for example, by telling the user what experiments to perform next to gain maximum information from minimum effort; Propose novel products and processes that optimise target properties; Test candidate products or process changes, reducing the need for expensive experiment or simulation; Capture and share knowledge by providing company-standard models and tools. The toolset also helps with the deployment challenges many organisations face when they try to use machine learning in practice. Powerful machine learning techniques can be accessed by scientists, engineers, and analysts via a simple web browser interface, while data science teams can integrate its algorithms into existing tools and workflows. The current and potential applications for the technology include: ● Design and optimisation of formulated products such as specialty chemicals, foods and beverages, inks, dyes, paints, and cosmetics. ● Design of new materials and optimisation of related processes - e.g., metal alloys, ceramics, plastics, surface treatments. ● Drug discovery - for example, guiding experimental programs to identify compounds of interest. ● Additive Manufacturing - exploring critical property / process relationships and enabling data-driven decisions about AM processes. ● Manufactured products - supporting informed decisions about product choices and process improvements and enabling predictive maintenance. ● Data science – the methods can be applied to extract value from any numerical or categorical dataset. For interested parties, a number of resources are available on the company’s website including white papers, academic papers, case studies and webinars The company is interested in commercial agreements with technical assistance, and technical cooperation incl under European projects. The partners would typically have a need for a new material, chemical or an improvement of a process where sparsity and inhomogeneity of data is a problem.