data_types Package

data_types Package

Signal processing data types (time series, feature vector, prediction vector)

According to the conventions given by MDP, all data types are two dimensional numpy arrays.

The data is exchanged between the nodes> via the node_chain module. For loading there is need of a source, for saving a sink node. For loading and saving these build the interface to datasets, where the data is collected and saved in certain datasets or loaded.

The BaseData type plays a special role, because it is a superclass for every other data type. The main role of this module is to equip each data type with the ability to store information beyond each node (if this is intended), i.e. it circumvents the original philosophy of the MDP framework. This behaviour is invisible for the user as long as it is not used.

Modules

base Type superclass providing common variables that survive the data processing
feature_vector Extend array of feature values with some additional properties
prediction_vector 1d array of prediction values with properties (labels, reference to the predictor)
time_series 2d array of channels x time series for windowed time series