Inference engine for efficient machine learning

A method at an inference engine comprising:

• executing a plurality of ordered steps to carry out inference on the basis of observed data;
• for each step, receiving a plurality of inputs to the step, and computing a predicted output and uncertainty of the prediction using a predictor;
• for each step, selecting either the predicted output or a known output on the basis of the uncertainty, and, if the known output is selected, computing the known output and retraining the predictor using the known output and the plurality of inputs of the step as training data.

One or more tangible device-readable media with device-executable instructions that, when executed by a computing system, direct the computing system to perform steps comprising:

• executing a plurality of ordered steps to carry out inference on the basis of observed data;
• for an individual step, receiving a plurality of inputs to the step, and computing a predicted output and uncertainty of the prediction using a predictor;
• for the individual step, selecting either the predicted output or a known output on the basis of the uncertainty, and, if the known output is selected, computing the known output and retraining the predictor using the known output and the plurality of inputs of the step as training data.