Data Transformers¶
Data transformers apply transformations to data. Note the transformations are limited to simple, in-place operations that does not change the shape of the data. If more complicated transformations like random projection or feature mapping are needed, consider using a data transformation layer instead.
- class DataTransformers.SubMean¶
Subtract mean from the data. The transformer does not have enough information to compute the data mean, thus the mean should be computed in advance.
- mean_blob¶
Default NullBlob(). A blob containing the mean data.
- mean_file¶
Default “”. When mean_blob is a NullBlob, this could be used to specify a HDF5 file containing the mean. The mean should be stored with the name mean in the HDF5 file.