The booklet gives a technical insight into the methods of machine learning and the main frameworks, but also ethics and law play a role.
While politicians are currently discussing the right way to deal with and possible support for artificial intelligence (AI), the methods of machine learning (ML) have also reached the point of productive use in Germany as well. Among other things, they help with fraud detection, the forecasting of machine failures and the evaluation of texts. Image analysis belongs to the paradigm discipline and supports among other things physicians in the diagnosis.
The AI reports tend to be one of two extremes: jubilant headlines or scare tactics. With this year’s iX-Developer Special Issue “Machine Learning”, heise developer opts for a factual and, above all, profoundly technical alternative. The 156-page booklet shows what can be achieved with ML methods and gives instructions for implementation.
Machine Learning does not do tasks on its own. Despite the opportunities, the challenges faced by developers and researchers are enormous. How do you create a machine learning model and see if it fits your purpose? What can be meaningfully implemented with artificial intelligence? And how do you check the result, which is no longer made up of algorithms that can be debugged – how do you look into the black box? The special issue wants to answer these questions.
From methods to frameworks to ethical questions
It covers the basics of ML methods like Supervised, Unsupervised and Reinforcement Learning as well as Artificial Neural Networks including Deep Neural Networks. An overview will help you choose the right open source framework, and articles with many code examples will help you get started with TensorFlow and Keras. Further contributions explain the advantages and disadvantages of the most important programming languages in the ML environment: from the top dog Python to the still set in the embedded environment C ++ to the functional challenger Scala. Extensive practical articles show the potential of machine learning, among other things in the image and text analysis for own projects.
A comparison of cloud offerings is intended for those who want to use pre-built ML services. It also gives readers insight into the hardware from using GPUs to special AI chips. Last but not least, the booklet deals with the important legal and ethical questions that the topic entails.
The special issue is from now until 13 December postage free orderable in the heise shop and in well-stocked newsagents available. The digital edition is available in the heise shop, Amazon, Weltbild, Bol and Beam.