Graphics from Aron Visuals –

Machine Learning - Sounds interesting, but whats the matter?

Machine Learning is taking an increasingly important part in our lives. This post will explain why is likely a relevant concept in the future.

Henrik Bartsch

Henrik Bartsch


For some years now, the term Artificial Intelligence has played an increasingly important role in news coverage, social media and everyday life. Intelligent systems in transportation, healthcare, finance are already commonplace. 1 Statistics show that the market behind artificial intelligence will continue to grow, making the topic likely to have a greater impact on everyday life in the coming years. 2 3 One term that keeps coming up is machine learning. But what is this actually all about?

Definition of machine mearning

Machine Learning is a subfield of artificial intelligence. It is the dominant subtopic of artificial intelligence, which is based on a mathematical-statistical basis and has already proven to be a successful concept. 4 5

In relation to artificial intelligence there are no meaningful demarcations to machine learning; often both terms are used synonymously.

Author’s comment: From my research, artificial intelligence is said to try to represent all the characteristics of humans at the same time. This is a characteristic that I do not yet see in today’s machine learning models - there we rather see models that are very good - sometimes even better than a human - in a certain delimited subdomain (also called weak AI).

Machine learning has shown to be a strong tool in many areas, as one of its strengths is the ability to find repetitive structures, regularities and relationships from data and experience - without being explicitly programmed with these relationships. Such algorithms are subsequently able to classify data, predict outcomes, or make informed decisions. 6 7

Machine learning can be divided into each of four different approaches in its conceptualization: 5 6

  1. Supervised learning: Learning an assignment of input data based on known output data,
  2. Unsupervised learning: Learning an assignment of input data based on unknown output data,
  3. Semi supervised learning: Learning an assignment of input data based on partially known and unknown output data,
  4. Reinforcement learning: Process optimization and systems control.

These different subdivisions have many different approaches, which come with different advantages and disadvantages with each other. Often similarities are recognizable.

Deep Learning

While machine learning is often spoken of, sometimes the term deep learning is also used, in parts also as a synonym. However, there are a few subtle differences here, which are important in the terminology. While machine learning basically includes any algorithms for learning from experience or data, deep learning only includes algorithms that are based on artificial neural networks. Such systems are based on abstracting a representation of the (human) brain and performing appropriate optimizations so that tasks can be solved appropriately.

Deep learning continues to differ from many other machine learning algorithms in that it is more flexible with respect to the data; often artificial neural networks are better able to independently find a wide variety of properties in the data. This eliminates work in the sense of pre-processing by humans. 8 6

Function fields of machine learning

There are with the machine learning three different application fields, into which the use can be divided: 9

  1. Descriptive analytics: From historical data it is tried to explain, what happened in the past.
  2. Predictive analytics: From historical data it is tried to explain what will happen in the future.
  3. Prescriptive analytics: From historical data, an attempt is made to generate future actions, which should optimally be used. should be used.

The term historical data describes data that has been generated and stored in the past.

This wide range of application fields enables machine learning to bring improvements or optimize processes in many areas of current life to bring improvements or to optimize processes in many areas of current life.

Fields of application

In addition to the fields of application mentioned at the beginning, some more fields of application can be found: 8 7

  1. speech recognition
  2. automation of customer service
  3. nature conservation, for example in protecting endangered species
  4. speech filtering, for example real-time filtering for toxic communications

For many more examples, see here.







  5. 2

  6. 2 3

  7. 2

  8. 2