Enhanced Random Processes
Version 10 expands on the already extensive random process framework with new processes, including hidden Markov models. Hidden Markov models are typically used to infer the hidden internal state from emissions, as in communication decoding, speech recognition, and biological sequence analysis. The random process framework also adds advanced time series processes and transformations of existing processes, as well as significantly improves computation with slice distributions—the bridge from random processes to random variables—often giving definite conclusions about expected process behavior from models.
- Support for scalar- and vector-valued hidden Markov processes. »
- Support for hidden Markov processes with discrete or continuous emissions.
- Support for hidden Markov processes with silent states.
- Find the sequence of hidden states from emissions using Viterbi and other decoding methods. »
- Automatically estimate hidden Markov process parameters from data.
- Build new processes as transformations of other processes. »
- Support for white non-Gaussian noise process. »
- Support for colored Gaussian noise process.
- Support for serial autocorrelation test of time series. »
- Substantially improved support for computation with time slices of processes across all random processes.
- Substantially improved simulation performance for most random processes.
- Substantial robustness and performance improvements of parameter estimation for many processes.