Harnessing the potential of data with machine learning

Through the targeted use of Machine learning process data can be used to gain new insights and make predictions. This enables companies to optimize or even completely redesign their business processes, from marketing and sales to after sales.

What opportunities and risks does machine learning bring for your business model?
Which current technologies are already being used successfully on the market?
Which best practices can you transfer to your operational business?

We discussed these and other questions with customers and partners at this year's IoT slashTalk spoken.

Just how quickly the limits of human cognitive perception are reached became clear during the presentation by doubleSlash IoT expert Simon Noggler. Like him, you would probably assume that you were looking at the logo of the lifestyle label "Hollister" in this picture.

However, a direct comparison shows that this is not the logo of the "Hollister" brand.

What happened here? The search engine was "fed" with a large number of images and, based on these Training data learned what which logo looks like. The search engine system did not make the human error of perception and recognized the seagull directly as the logo of the "Cleptomanicx" brand.

So does the computer beat humans when it comes to learning?

Machine learning in a mobile and networked world - opportunities and challenges

Dr. Dirk Wacker, Director Technology and Innovation at Giesecke+Devrient Mobile Security GmbH, gave a Insight into the broad field of machine learning - From the computer game Alpha Go to self-service offerings in medicine and autonomous driving - machine intelligence is now used in many applications. His key message: Machine learning - a gift and a curse.

Democratization of machine learning - paths to practical implementation

Nico Wilhelm, Partner Development Manager at Microsoft Deutschland GmbH and Dr. Philipp Kesten, Senior Director Field Engineering at PTC talked about why Cloud and infrastructure so important for machine learning projects and that one of the most important prerequisites for the democratization of technology is the ease of use for specialist departments.

Who evaluates better: man or machine? Application of machine learning to an operator-based laboratory test

Dr. Thomas Mühlenstädt, Industrial Statistician at W. L. GORE & Associates GmbH, provided insights into the Test procedure via image recognition for Gore-Tex® materials and described the challenges and opportunities inherent in this process.

The slashTalk was rounded off with three lighning talks by our doubleSlash colleagues on the topic of machine learning.

Ralf Richter, IoT Lead Developer, addressed the question of why AI loves the cloud:

Data integration expert Markus Beller gave an insight into the error-proneness of image recognition algorithms:

Julian Mehne, expert for Data Scienceshowed how biases can influence machine learning:

Do not neglect change management

The speakers' insights and best practices have shown that Change management is also a very important task in machine learning projects. Because when human tasks are replaced by machines, questions and possibly even fears naturally arise among employees. These should be dealt with sensitively. At the same time, machine learning brings with it new requirements. So the Requirement profiles for employees - away from being an operator and towards being a supervisor and further developer of the system. This needs to be considered in the project and this change process needs to be accompanied.

Biases in machine learning projects

The effects of Biases in machine learning projects should not be underestimated, as "garbage in - garbage out" also applies here. If the data (input) is not clean or not representative, then the informative value of the model (output) is also not satisfactory.
The selection of the tool, e.g. a particular Algorithmimplies a certain bias in itself. This can lead to unsatisfactory results.

Learn more about machine learning