dataset_defs Package¶
dataset_defs
Package¶
Store and load accumulations of datasets together with meta data
Datasets are normally several data sets arranged in several folders with some additional meta data, giving information, e.g. about origin and type. The type specifies witch dataset class should handle the data. When writing new datasets it is important to follow this naming rules:
- type and module name of the new dataset in lower-case with underscores to separate words, e.g. time_series and time_series.py
- class name of the new dataset in camel-case and with ending ‘Dataset’, e.g. TimeSeriesDataset
When stored as results of operations or operation chains, datasets already have the needed format, but some datasets are also able to transform data to the needed format.
Nevertheless, you will need a metadata.yaml file. For more details have a look at: The Data Directory (storage).
Datasets are mostly stored as results of an
operation
.
Modules¶
base |
Base Module for datasets to specify the interface for these |
dummy |
Store only meta data but no real data (except from store state of nodes) |
feature_vector |
Load and store data sets containing Feature Vectors |
metric |
Methods to calculate and store classification results (metrics) |
performance_result |
Tabular listing data sets, parameters and a huge number of performance metrics |
prediction_vector |
Load and store data sets containing Prediction Vectors |
stream |
Reader objects and main class for continuous data (time series) |
time_series |
Load and store data of the type pySPACE.resources.data_types.time_series |