Shannon's entropy can be used to measure Mutual Information between input and output variables. Also to detect outliers. I have worked on that topics but I don't know how it scales to Big Data...
Mutual information for the selection of relevant variables in spectrometric nonlinear modelling
Shannon's entropy can be used to measure Mutual Information between input and output variables. Also to detect outliers. I have worked on that topics but I don't know how it scales to Big Data...
Mutual information for the selection of relevant variables in spectrometric nonlinear modelling
entropy can be used as one of the features during machine learning (I have seen many papers using it, sometimes I also use it), but it depends on particular problem whether it will bring any add value
Shannon entropy has been used extensively to measure the complexity of finite graphs, and has been used in chemical graph theory for purposes of classification. A typical procedure is to determine a partition of graph elements (e.g., vertices or edges), construct a finite probability scheme in the obvious way, and apply Shannon's entropy function to the probability scheme. The key is determining a partition (relative to some structural property such as symmetry or path lengths), often a computationally hard problem. Big data would be more challenging unless there is some way of approximating a partition of graph elements.