Shota Saito, Yuta Nakahara, Toshiyasu Matsushima
The Soft-BCT model introduces a probabilistic approach to context tree models for real-valued time series, showing competitive performance with existing models.
This study introduces a new model called the Soft Bayesian Context Tree (Soft-BCT) for analyzing time series data, which are sequences of data points collected over time. Unlike previous models that make strict decisions about how to split the data, Soft-BCT uses a probabilistic method, allowing for more flexible analysis. The researchers developed a learning algorithm using a statistical technique called variational inference. When tested on real-world data, the Soft-BCT performed as well as or better than older models.
