**
A useful machine-learning application would be to find an "executive summary" of a time series.
By this I mean another time series, which when subtracted from the original series yields a
series that can be captured by a simple statistical model (e.g. a normal variate). The utility of
such a "tendency" is that it would encapsulate the most critical aspects a dynamicist might need
to explain if this signal has a rational basis. The intrinsic time decomposition is shown to be a
diffusion filter that can handle multiscale dynamics. A "tendency" can be produced from an
intrinsic time decomposition, followed by a selection criterion that favors empirical statistical
symmetries and information content. The tendency, when applied to a climate signal, will show
the warming trend as well as the preponderance of the ocean in controlling the time scales of
fluctuations.
**

**
**

To download: Right-click and choose "Save Link As..." (Other video options)

To begin viewing slides, click on the first slide below. (Or, view as pdf.)