Universal Time-Series Forecasting with Mixture Predictors

The author considers the problem of sequential probability forecasting in the most general setting, where the observed data may exhibit an arbitrary form of stochastic dependence.