ZScoring¶
Produce z-scored data, that is, data with a mean of zero and a variance of one.
The resulting data can be interpreted to be scores in units of standard deviations (Z units), relative to the sample mean. Note that, if your data is segmented or otherwise comes in the form of multiple instances, you can decide whether the statistics should be taken over the instances or over the time axis in each instance. You also can decide whether the statistics should be updated on every chunk, or if it should only be updated on the calibration data (non-streaming chunks), and whether they should be recomputed on subsequent non-streaming chunks (e.g., subsequent test data). Note that there is also a moving-window standardization node available. More Info... Version 1.2.0
Ports/Properties¶
data¶
Data to process.
axis¶
Axis over which the statistic should be aggregated. For instance, if this is set to time, then the mean/standard deviation will be computed over time.
pop_correction¶
Use Bessel's correction for the population standard deviation. If True, the standard deviation will be scaled by 1/(N-1), which yields a less biased estimate of the standard deviation of the (infinite) population from which the sample is assumed to be drawn, while if False, it will be scaled by 1/N, which yields the maximum-likelihood standard deviation estimate for the (finite) sample. Note that this is not applicable (and therefore ignored) in the robust mode, where the median absolute deviation is used instead of the standard deviation.
robust¶
Use robust estimators (median and median absolute deviation). Enabling this will make the node more robust to outlier elements, but if the data is clean, then the statistic will be somewhat less precise than it it could be.
adapt_on_streaming¶
Adapt state parameters on streaming chunks. If this is False (the default), then this node will not adapt itself on streaming chunks, i.e., only on non-streaming chunks, such as imported calibration recordings. This is useful if the streaming chunks would be too short to meaningfully adapt the parameters on and the calibration recording is sufficietly similar to the streaming data for it to be meaningful to adapt the state only on these data. Or if set to True, the node will simply adapt itself on every chunk it receives, whether it is streaming or not.
constrain¶
Optionally constrain data to have the same scale across one or more axes. Multiple axes can be given as a comma-separated list. If 2 axes are given, rank-1 offset and scale matrices are estimated using SVD and NMF, respectively.
initialize_once¶
Initialize (adapt state parameters) only once. If False, this node will recalibrate itself on any qualifying chunk (based on the setting of adapt on streaming chunks).
ignore_resets¶
Ignore signal resets.
ignore_nans¶
If activated the data containing nan values are ignored.
epsilon¶
Optional epsilon value to prevent div by 0.
partition_size¶
Partition size. Optionally a size of chunks into which data shall be partitioned to conserve memory (if robust=True). This is done along the last axis of the data that isn't the axis of interest. This is a rarely used feature.
backend¶
Compute backend to use.
precision¶
Numeric precision to use. Can be reduced to save memory (e.g. if running on GPU).
set_breakpoint¶
Set a breakpoint on this node. If this is enabled, your debugger (if one is attached) will trigger a breakpoint.
metadata¶
User-definable meta-data associated with the node. Usually reserved for technical purposes.