Comparative study of visualisation methods for temporal data

In this paper, we investigate a financial data set from [5] using two algorithms which are both designed for visualising data. One algorithm consists of a neuroscale algorithm which uses different Bregman divergences. The other uses a similar algorithm but based on reservoir computing. We show that the latter is much better because it captures the dynamical nature of the financial time series and thus reveals more explicit information. By investigating a slightly different cost function, we show that the latter mapping is not creating information which does not exist in the original time series.