Explanation of terms and concepts
| Artificial intelligence (AI) | General term when computer systems take over tasks that are typically assigned to human attributes such as learning, recognising, planning and so on. Can also be robots or cars that move independently in their environment. |
| Algorithm | Set of steps for a computer program to accomplish a task or to solve a problem. |
| Machine learning (ML) | Subform of AI. Computer systems that learn and adapt independently from data without following explicit instructions. Can be prediction models or image recognition. |
| Supervised learning (SL) | Subform of ML. Models are trained and validated in existing, labelled data sets. These are typically used for classification tasks, for example, to predict future disease states or to detect pathologies on images. |
| Unsupervised learning (UL) | Subform of ML. Models are created from unlabelled data, for example, for clustering or outlier detection in electronic medical records. |
| Reinforcement learning (RL) | Subform of ML. Models that can make prospective decisions on their own and constantly improve them depending on the results. Works through a reward function (trial and error). Only good actions continue. |
| Q-learning | Subform of RL. A model-free, flexible RL algorithm to learn the value of a certain action. Random actions outside a specific system can be learnt, for example, by imitating and improving expert actions. |
| Artificial neural networks | A set of algorithms, modelled loosely after the human brain, in the form of different layers similar to neurons. A powerful tool which can be used for supervised, unsupervised or RL. |