Project SEKO#

iplus1 regularly implements end-to-end machine learning systems that are used by very different types of users. Sometimes those are IT or even data science professionals that know how to use, parametrize and evaluate the results of data pipelines and algorithms.

There are however a far bigger class of users that are experts and professional in another domain who know a great deal about the processes, measurements and outcomes of their domain but have little to no experience with data science or machine learning.

Self-service machine learning for domain experts#

Given those users the tools to set up, use and improve a complete machine learning system in a “self-service” fashion is a great challenge. The SEKO project aims to tackle this challenge.

For example, in a setting where a system’s future state should be predicted and supervised machine learning algorithms are useful, there typically a dearth of labeled data. Expecting domain experts to label large amounts of data is unrealistic.

Fraunhofer has developed and patented an machine learning algorithm called COMETH that works on minimal amounts of labeled data. Even more crucial, it can detect where in the data additional labels would be most useful. This can be used to guide a human expert to label the most informative data points.

The SEKO project evaluates if and how this can be used to reduce the amount of needed expertise and effort to set up COMETH so that domain experts can do it themselves.

We are looking forward to another great joint project of iplus1 and Fraunhofer IOSB-INA!