Data-driven solutions based on machine and deep learning are gaining momentum in the automotive industry beyond autonomous driving. Josef Viehhauser and Dominik Schniertshauer explore use cases from the BMW Group where novel machine-learning pipelines (such as those based on XGBoost and convolutional neural nets, for example) support a broad variety of business stakeholders. These have proven to be effective in and have created value for domains including vehicle engineering and aftersales. Along the way, Josef and Dominik share best practices and lessons learned from both an architectural and methodological perspective.
Josef and Dominik begin by describing the application of CNNs in the context of computer vision use cases in manufacturing processes before discussing XGBoost and other methods for predicting the residual value of leased vehicles. (One of the challenges in such use cases is the lack of training data.) They conclude with a survey of several possibilities that help deploy models created by machine learning in productive IT setups and look at the future of “end-to-end machine learning.”
Josef Viehhauser is a full stack data scientist at the BMW Group, where he leverages machine learning to create data-driven applications and improve established workflows along the company’s value chain. Josef also works on scoping and implementing such use cases in scalable ecosystems primarily via Python. Outside of work, he is interested in technological innovations and soccer.
Dominik Schniertshauer is a data scientist at the global headquarters of the BMW Group in Munich, Germany. As a deep learning enthusiast, Dominik dedicates himself to solving complex problems in the context of customer, logistics, and production data, focusing on the end-to-end character of deep learning use cases and their scalable implementation.
©2017, O’Reilly UK Ltd • (800) 889-8969 or (707) 827-7019 • Monday-Friday 7:30am-5pm PT • All trademarks and registered trademarks appearing on oreilly.com are the property of their respective owners. • email@example.com