![]() |
ИСТИНА |
Войти в систему Регистрация |
ИПМех РАН |
||
We propose a technique and a particular machine learning model architecture that helps researchers in different domains to automatically and explicitly reveal functional patterns in multivariate time series from a series of experiments. We formalize the original task in terms of machine learning, and there is no need to explore the subject domain. The proposed architecture uses information about the recording device. The initial position of the recording electrodes is encoded as a graph and transmitted to the corresponding architecture. The efficiency of the technique has been demonstrated in the field of neurophysiology for the data where the P300 pattern is already known to exist. As to further research, it is of interest to extend the proposed technique on other domains, for example, data from sensors on production lines or banking transactions.