Discrete Sequence Prediction and Its Applications

摘要: Learning from experience to predict sequences of discrete symbols is a fundamental problem in machine learning with many applications. We present simple and practical algorithm (TDAG) for sequence prediction. Based on text-compression method, the TDAG limits growth storage by retaining most likely prediction contexts discarding (forgetting) less ones. The storage/speed tradeoffs are parameterized so that can be used variety Our experiments verify its performance data compression tasks show how it applies two problems: dynamically optimizing Prolog programs good average-case behavior maintaining cache database mass storage.