A lot of software was developed for the VAX and subsequently the Alpha platform.
The status of the good old machine can be checked at
Most programs are not readily
portable. Source code, or a login account on RULLF2 can be requested from the
- Short-term forecasting is performed for a set of (test) delay
vectors by constructing piecewise-linear approximations to
trajectories in reconstructed phase space. The model contains
information from nearest neighbours in a second (fitting) set
of delay vectors. Small correspond to fitting deterministic
models, intermediate to nonlinear stochastic, and largest
to fitting a linear stochastic autoregressive model.
- The implemented DVS ("Deterministic Versus Stochastic") algorithm is described
in [Casdagli and Weigend, 1993].
- Fig. 1 shows the results for a time series of paroxysmal atrial
fibrillation (type I) in a patient.
- Calculation of the Integral Local Deformation (ILD) and Fill Factor
(FF). These parameters may be used to determine an optimal embedding
delay and embedding dimension for reconstructing phase space with
delay vectors. The Mutual Information (MI) is also calculated.
- The implemented algorithms are described
in [Buzug et al., 1990,Buzug and Pfister, 1992a,Buzug and Pfister, 1992b].
- Fig. 2 shows the results for the y-component of the
Lorenz system [Buzug and Pfister, 1992b].
- MLDK2 and MLDK2FIT
- Provide various maximum likelihood estimators of the correlation
dimension and entropy.
- The implemented algorithms are described in [Olofsen et al., 1992] and
- Figures 3 and 4 show the results of
the analysis of a recording of atrial fibrillation [Olofsen et al., 1992].