As long as it is annotated data, STOCHOS can be used to carry out sample- and resource-efficient, expert-independent probabilistic machine learning, using existing data or to plan new samples. It is (almost) irrelevant whether the data comes from experiments or simulations. In addition to this supervised learning (normally used for sensitivity analysis and optimization), STOCHOS is also able to carry out unsupervised learning, for example to carry out predictions of time series in order to then carry out optimal control using reinforcement learning. Since STOCHOS can be used equally on CPU/GPU/TPU and even runs on iOS/Android/Linux Embedded/Microcontroller, “real-time” response times are possible.