1.1 What is MLPro?

MLPro is a synoptic and standardized Python package to produce a solution for standard Machine Learning (ML) tasks. In the first version of MLPro, MLPro provides sub-packages for a subtopic of ML, namely Reinforcement Learning (RL), which is developed under a uniform infrastructure of basic and cross-sectional functionalities. MLPro supports simulation as well as real-hardware implementations. MLPro team has developed this framework by taking care of several main features, such as CI/CD method, clean code, object-oriented programming, ready-to-use functionalities, and clear documentation.

Additionally, we use established and well-known scientific terminologies in the naming of the development objects. Although MLPro is standardized and has a high complexity, we make the implementation of MLPro as easy as possible, understandable, and flexible at the same time. One of the properties of being flexible is the possibility to incorporate the widely-used third party packages in MLPro via wrapper classes. The comprehensive and clear documentation also helps the user to quickly understand MLPro.

One of the main advantages of MLPro is the complete structure of MLPro that is not limited to only environments or policy and is not restricted to any dependencies. MLPro covers environment, agents, multi-agents, model-based RL, and many more in a sub-framework, including cooperative Game Theoretical approach to solve RL problems.

We are committed to continuously enhancing MLPro, thus it can have more features and be applicable in more ML tasks.

Key Features

• Numerous extensive sub-frameworks for relevant ML areas like reinforcement learning, game theory, online machine learning, etc.

• Powerful substructure of overarching basic functionalities for mathematics, data management, training and tuning of ML models, and much more

• Numerous wrapper classes to integrate 3rd party packages