Graph-based Predictable Feature Analysis
This work addresses the challenge of extracting predictable patterns in time series data for applications in fields like signal processing or machine learning, but it appears incremental as it builds on and compares to prior methods.
The authors tackled the problem of unsupervised learning of predictable features from high-dimensional time series by proposing graph-based predictable feature analysis (GPFA), which defines predictability as low variance in future data given past data, and showed competitive results compared to existing methods like slow feature analysis and forecastable component analysis.
We propose graph-based predictable feature analysis (GPFA), a new method for unsupervised learning of predictable features from high-dimensional time series, where high predictability is understood very generically as low variance in the distribution of the next data point given the previous ones. We show how this measure of predictability can be understood in terms of graph embedding as well as how it relates to the information-theoretic measure of predictive information in special cases. We confirm the effectiveness of GPFA on different datasets, comparing it to three existing algorithms with similar objectives---namely slow feature analysis, forecastable component analysis, and predictable feature analysis---to which GPFA shows very competitive results.