Machine learning for IT Services

With machine learning IT becomes suitable for everyday use. IT used to be defined by deterministic processes and procedures, but machine learning allows large amounts of unstructured data to be processed. In other words: The data that arises in everyday life no longer has to be specially prepared, but instead can be fed directly into an IT process.

Unstructured data relates to, for example, documents or the text in an input field on a web form, but also images or sounds. Texts, pictures and sounds can thus be enjoyed as such. Often you want to “do something” with it, and for this purpose this data must be categorised. In the past this work had to be done by people, and this process was quite laborious. Machine learning replaces this tedious work and not only makes it significantly faster, but often also much more precise.

In IT Services, we want to offer first-class services to our customers at ETH. Our customers in the departments and sections are often faced with a lot of unstructured data. It is therefore time for us to acquire the necessary knowledge in the field of machine learning, so that we can provide help in such cases. These people should be doing research or providing their personal services and not wasting their time on categorisation work.


At the beginning of summer, the sections ITS SIS and ITS SWS held two three-day workshops for this purpose. In Tim Head ( we have found an outstanding Workshop Head. The first workshop was geared more towards the needs of SWS software developers and focused on the processing of texts. During this workshop they learned about spaCy and how to use Keras, for example. SpaCy is a Python library for natural language processing (NLP). Keras is also written in Python and can be used to model neural networks in order to exploit deep learning.

As the term “deep learning” has appeared more frequently in the scientific press in recent years, knowledge in this area is increasingly being requested from Scientific IT Services. For example, the recognition of cells in microscope images and so-called image segmentation plays an important role for customers in the biology area, and the latest methods make use of deep learning. Cosmologists at ETH Zurich also try to determine the characteristics of galaxies with machine learning using images. Therefore, the second workshop focused on the processing of image data.

In addition to an introduction to the basics, we have also experimented with neural networks to classify handwritten numbers and images of garments. In doing so it became clear that deep learning does not work satisfactorily out-of-the-box, and that sometimes time-consuming testing of network architectures is required to achieve good results.

We are certain that machine learning will play an increasingly important role for our customers. This workshop was the first step with which we have started to establish and expand our expertise in this area in the sections ITS SWS and ITS SIS. We are thus prepared to offer our services quickly and professionally, if necessary.

Text and contact

Benno Luthiger, ITS WCMS, LMS & Mobile Applications, Software Services (ITS SWS)
Uwe Schmitt, ITS Scientific Software & Data Mgmt., ITS Scientific IT Services (ITS SIS)

erstellt am
in News Schlagwörter:

Hinterlassen Sie eine Antwort

Ihre E-Mail-Adresse wird nicht veröffentlicht. Erforderliche Felder sind mit * markiert

Diese Website verwendet Akismet, um Spam zu reduzieren. Erfahren Sie mehr darüber, wie Ihre Kommentardaten verarbeitet werden .