Reference Manual¶
As seen, the ingredients for validating a model are the model itself, a dataset and some validation metrics. However, Dymoval is not a modeling tool and therefore its focus areas are the following
Dymoval Architecture¶
Dymoval has been thought to help engineers in analyzing datasets and in developing models by providing a validation tool.
Although there plenty of amazing packages out there like pandas, numpy, matplotlib, etc., they are huge and the plethora of functionalities they offer may be overwhelming.
Therefore, the idea is to combine the functionalities of the aforementioned tools in such a way engineers are not overwhelmed but at the same time they can manage to get their job done quickly and effectively. However, we do not want to limit engineers, but we want to guarantee access to the tools Dymoval has been built upon in a seamless way.
Therefore, Dymoval classes are composed as shown in the picture below.
Dymoval structure. You can access every attributes and methods of an inner object from an outer object.¶
This means that every outer package/module have full access to the classes/functions provided by the inner packages/modules. For example, it is possible to access all Dataset methods from a ValidationSession object and all the pandas classes and methods from Dataset objects.
Furthermore, object methods are not “inplace” but they always return a modified version of the calling object. For example
>>> ds.remove_means() # won't change ds
>>> ds = ds.remove_means() # you must re-assign ds
Finally, each plotting function returns a matplotlib figure so that the user can access all the matplotlib API for further manipulating the figure in case he/she is not happy with the results from Dymoval.
Warning
Given that a outer objects do not contain only inner object types attributes, and given that such attributes are connected to other attributes, it is discouraged to directly change inner class attributes with inner methods.
For example, we know that a Dataset object attribute is a pandas Dataframe, but if we change such a pandas DataFrame, then we shall update all the other Dataset attributes such as the coverage region, Nan intervals, etc. and that may become very messy. Therefore, if you want to change any attribute of any Dymoval object, then use the Dymoval class methods (if any) or create a new class instance.
Package structure¶
Dymoval’s package is arranged in the following modules
Module containing everything related to datasets. |
|
Module containing everything related to validation. |
|
Module containing some useful functions. |