No abstract

In this review, we present an analysis of the most used multi-agent reinforcement learning algorithms. Starting with the single-agent reinforcement learning algorithms, we focus on the most critical issues that must be taken into account in their extension to multi-agent scenarios. The analyzed algorithms were grouped according to their features. We present a detailed taxonomy of the main multi-agent approaches proposed in the literature, focusing on their related mathematical models. For each algorithm, we describe the possible application fields, while pointing out its pros and cons. The described multi-agent algorithms are compared in terms of the most important characteristics for multi-agent reinforcement learning applications—namely, nonstationarity, scalability, and observability. We also describe the most common benchmark environments used to evaluate the performances of the considered methods.

The evaluation of the performances of brain-computer interface (BCI) systems could be difficult as a standard procedure does not exist. In fact, every research team creates its own experimental protocol (different input signals, different trial structure, different output devices, etc.) and this makes systems comparison difficult. Moreover, the great question is whether these experiments can be extrapolated to real world applications or not. To overcome some intrinsic limitations of the most used criteria a new efficiency indicator will be described and used. Its main advantages are that it can predict with a high accuracy the performances of a whole system, a fact that can be used to successfully improve its behavior. Finally, simulations were performed to illustrate that the best system is built by tuning the transducer (TR) and the control interface (CI), which are the two main components of a BCI system, so that the best TR and the best CI do not exist but just the best combination of them.

No abstract