Processing sensor data locally requires considerably less energy than communicating it to a distant node, yielding an interesting communication / computation trade off.

To reduce global communication requirements, one needs to perform signal processing to extract key information in a distributed fashion and without losing fidelity.

We have considered the problem of distributed classification in a WSN by means of support vector machines (SVM).

Taking advantage of the sparse representation that SVMs provide for the decision boundaries, we designed classes of

• incremental and
• gossip-based

distributed consensus algorithms for training the classifier.