Neo imposes and recommends some attributes for all objects, and also provides the annotations dict for all objects to deal with any kind of extensions. This flexible feature allow Neo objects to be customized for many use cases.
While any names can be used for annotations, interoperability will be improved if there is some consistency in naming. Here we suggest some conventions for annotation names.
SpikeTrain.annotations[‘waveform_features’] : when spike sorting the waveform is reduced to a smaller dimensional space with PCA or wavelets. This attribute is the projected matrice. NxM (N spike number, M features number. KlustakwikIO supports this feature.