I want to apply machine learning to the problem of classification in a parallel environment. Several independent nodes, each with multiple on / off sensors, can transmit their sensor data to classify the event, as defined by heuristic, training data, or both.
Each peer will measure the same data from their unique point of view and will try to classify the result, taking into account that any neighboring node (or its sensors or just a connection to node) may be faulty. The nodes should function as equal peers and determine the most probable classification, reporting their results.
Ultimately, each node must make a decision based on its own sensor data and its peers. If that matters, false positives are suitable for certain classifications (albeit undesirable), but false negatives would be completely unacceptable.
Given that each final classification will receive good or bad feedback, what would be the appropriate machine learning algorithm to solve this problem if the nodes could communicate with each other to determine the most likely classification?
source
share