In an example, a network is described with a plurality of data sources. Each data source may provide a feature, such as a data type that the data source collects or generates. A data aggregator may be connected to the network, and configured to collect, classify, and merge features as appropriate. The data aggregator includes a discriminator for classifying features, a merger, unmerger, converter, and evaluator. Features are provided to one or more expert systems configured to control one or more systems based on the features. Feedback to the data aggregator is used to evaluate the success of a merge. When a merge is found to be unhelpful, features may be unmerged.