Data reduction for reducing a data set

A method of reducing a data set based on a subset of variables from a set of variables comprising a plurality of variables, the plurality of variables representing different types of data, instances of the plurality of variables being based on sensor data and comprising information to predict an instance of a further type of data, the further type of data relating to a characteristic of a physical entity, the method comprising:

• obtaining a first set of data comprising tuples of instances of data, wherein each tuple comprises an instance of each variable of the set of variables;
• obtaining a second set of data comprising instances of the further type of data, wherein each instance of the second set of data corresponds to one of the tuples of the first set of data;
• searching for a reduced set of variables being a subset of the set of variables, the reduced set of variables representing an at least local optimum of an optimization function being a function of a reduced first set of data, the second set of data and the reduced set of variables, the reduced first set of data comprising reduced tuples obtained from the tuples of the first set of data, and the reduced tuples only comprising instances of the variables of the reduced set of variables, the optimization function being based on a combination of a first mutual information value between the reduced first set of data and the second set of data and a penalty value that increases with an increasing number of variables in the reduced set of variables;
• and
• generating a reduced data set from a data set that comprises tuples of instances of data, wherein each tuple comprises an instance of each variable of the set of variables, the reduced data set comprising instances of data of variables that are present in the reduced set of variables.

A machine, comprising:

• a plurality of sensors configured to generate sensor data;
• a physical entity;
• and
• a processor to perform a method for reducing a data set based on a subset of variables from a set of variables comprising a plurality of variables, the plurality of variables representing different types of data, instances of the plurality of variables being based on sensor data and comprising information to predict an instance of a further type of data, the further type of data relating to a characteristic of the physical entity, the method comprising:
• obtaining a first set of data comprising tuples of instances of data, wherein each tuple comprises an instance of each variable of the set of variables;
• obtaining a second set of data comprising instances of the further type of data, wherein each instance of the second set of data corresponds to one of the tuples of the first set of data;
• searching for a reduced set of variables being a subset of the set of variables, the reduced set of variables representing an at least local optimum of an optimization function being a function of a reduced first set of data, the second set of data and the reduced set of variables, the reduced first set of data comprising reduced tuples obtained from the tuples of the first set of data, and the reduced tuples only comprising instances of the variables of the reduced set of variables, the optimization function being based on a combination of a first mutual information value between the reduced first set of data and the second set of data and a penalty value that increases with an increasing number of variables in the reduced set of variables;
• and
• generating a reduced data set from a data set that comprises tuples of instances of data, wherein each tuple comprises an instance of each variable of the set of variables, the reduced data set comprising instances of data of variables that are present in the reduced set of variables.

A non-transitory computer-readable medium having one or more executable instructions stored thereon, which when executed by a processor, cause the processor to perform a method for reducing a data set based on a subset of variables from a set of variables comprising a plurality of variables, the plurality of variables representing different types of data, instances of the plurality of variables being based on sensor data and comprising information to predict an instance of a further type of data, the further type of data relating to a characteristic of the physical entity, the method comprising:

• obtaining a first set of data comprising tuples of instances of data, wherein each tuple comprises an instance of each variable of the set of variables;
• obtaining a second set of data comprising instances of the further type of data, wherein each instance of the second set of data corresponds to one of the tuples of the first set of data;
• searching for a reduced set of variables being a subset of the set of variables, the reduced set of variables representing an at least local optimum of an optimization function being a function of a reduced first set of data, the second set of data and the reduced set of variables, the reduced first set of data comprising reduced tuples obtained from the tuples of the first set of data, and the reduced tuples only comprising instances of the variables of the reduced set of variables, the optimization function being based on a combination of a first mutual information value between the reduced first set of data and the second set of data and a penalty value that increases with an increasing number of variables in the reduced set of variables;
• and
• generating a reduced data set from a data set that comprises tuples of instances of data, wherein each tuple comprises an instance of each variable of the set of variables, the reduced data set comprising instances of data of variables that are present in the reduced set of variables.