3.1. Overview

MLPro has an extensive substructure of comprehensive basic functionalities, which are combined in the sub-framework MLPro-BF. This is organized in a total of five layers that build on one another, as shown in the following figure:

The lowest Layer 0: Elementary Functions provides a collection of functions for logging, time measurement in simulated or real processes, persistence and data management, plotting, etc. It also contains a framework for interactive GUI applications.

The next Layer 1: Computation consists of various functionalities related to computation topics, such as event handling and multitasking. Furthermore, it introduces an abstract runtime scenario with an operation mode (simulation or real operation). This is one of the key concepts in MLPro to support real applications. It is reused and specialized at higher levels.

On top of this, Layer 2: Mathematics introduces elementary mathematical objects like dimensions, sets and elements, metric spaces, and functions. Furthermore, numeric algorithms for data normalization etc. are included.

Layer 3: Application Support prepares the connection to real applications. It introduces powerful systematics for stream data processing/visualization and state-based systems that are, in turn, prepared for communication with real hardware components like sensors and actuators.

The top Layer 4: Machine Learning of MLPro-BF specifies fundamental standards for machine learning. All higher ML-related sub-frameworks reuse and specialize them. Topics like hyperparameters, adaptive models, and their training and tuning in ML scenarios are handled here.