1.1 Machine Learning
1.1.1 What is Machine Learning?
An algorithm is a well-defined sequence of instructions that, for a given sequence of input arguments, yields some desired output.
In other words, it is a specific recipe for a function.
Developing algorithms is a tedious task.
In machine learning, we build and study computer algorithms that make predictions or decisions but which are not manually programmed.
Learning needs some material based upon which new knowledge is to be acquired.
In other words, we need data.
1.1.2 Main Types of Machine Learning Problems
Machine Learning Problems include, but are not limited to:
Supervised learning – for every input point (e.g., a photo) there is an associated desired output (e.g., whether it depicts a crosswalk or how many cars can be seen on it)
Unsupervised learning – inputs are unlabelled, the aim is to discover the underlying structure in the data (e.g., automatically group customers w.r.t. common behavioural patterns)
Semi-supervised learning – some inputs are labelled, the others are not (definitely a cheaper scenario)
Reinforcement learning – learn to act based on a feedback given after the actual decision was made (e.g., learn to play The Witcher 7 by testing different hypotheses what to do to survive as long as possible)