The goal of the thesis is the survey of of deep reinforcement learning algorithms based on the Monte Carlo tree search (MCTS) framework. The ultimate focus is on a clear, readable, and thorough explanation of the Stochastic MuZero algorithm. The goal is for a reader knowledgeable in general reinforcement learning (and its implementation aspects) to be able to design a crude implementation of the aforementioned algorithms based on the information acquired from the thesis. Hence, the expected outputs of the thesis are as follows:

• Description and explanation of the Monte Carlo tree search framework
• Description and an explanation of the key MCTS-based reinforcement learning algorithms, namely AlphaZero, MuZero, and stochastic MuZero. This will include not only the explanation of the core MCTS-based components, but also a discussion of the neural net components, their architecture, and training.
• A lightweight experimental part which will report on an application of an existing implementation of (stochastic) MuZero on some standard benchmark of the author's choice. The goal is to report on possible pitfalls of applying the studied algorithms in practice.