The functions in TIM Matlab are highly similar. Therefore writing separate documentation for each function would be redundant and present maintenance issues. To avoid this, we have instead decided to factor those similarities into this generic documentation. In Matlab, this text is available by typing:
>> help tim
A signal is a real (m x n)-matrix that contains n samples of an m-dimensional signal.
Consider a signal as the result of a single experiment. When making such an experiment it is often useful to repeat it many times, perhaps to obtain lower variance for the measurement. These repetitions are called trials. If the different trials are comparable to each other, then TIM can make use of them more efficiently in the estimations, compared to computing estimates for the trials and then combining the results.
A signal set is an arbitrary-dimensional cell-array whose linearization contains the trials of a signal. A real array is interpreted as a cell-array containing one trial. The estimators take as parameters signal sets rather than single signals.
When an estimator takes as input multiple signal sets, then they are required to contain the same number of trials. The dimension of the signals must be equal inside a signal set, but need not be equal between signal sets. The number of samples may vary from signal to signal even inside a single signal set.
The signals can be applied time-delays in the estimators, and the amount of this delay is called a lag. Each lag can be given either as a scalar or as an array. If a lag is given as an array it means that the user wishes to repeat the estimation many times with the lag varying as specified in the array. Those functions that would output a scalar, then output a column vector, and those functions that would output a row vector, then output a matrix. In case some of the lags are given as arrays, those arrays must have the same number of elements, and a scalar lag is interpreted as an array of the same size with the given value as elements. The default for a lag is 0.
When using temporal versions of the estimators, the temporal adaptivity
is obtained by time-windowing the signals. In these estimators, the
timeWindowRadius parameter determines the radius of the time-window.
Another control on the time-adaptivity is given by the filter
parameter. It is an arbitrary-dimensional real array whose linearization
contains the coefficients by which to weight the results in the time-window.
The coefficient in the middle of the array corresponds to the current
time instant. The size of the array can be arbitrary but must be odd.
In addition, the coefficients must sum to a non-zero value.
Default [1].
Many of the estimators are based on k:th nearest neighbors. The
k here can be specified in many of the estimators. The default
for it is 1.