Prediction-Oriented Subsampling from Data Streams
This work addresses computational efficiency in learning from data streams, but it is incremental as it builds on existing information-theoretic approaches.
The paper tackles the challenge of intelligent data subsampling from streams for offline learning by proposing a prediction-oriented information-theoretic method, which outperforms a prior technique on two widely studied problems.
Data is often generated in streams, with new observations arriving over time. A key challenge for learning models from data streams is capturing relevant information while keeping computational costs manageable. We explore intelligent data subsampling for offline learning, and argue for an information-theoretic method centred on reducing uncertainty in downstream predictions of interest. Empirically, we demonstrate that this prediction-oriented approach performs better than a previously proposed information-theoretic technique on two widely studied problems. At the same time, we highlight that reliably achieving strong performance in practice requires careful model design.