Construct argument list from input data layout. #6
Milestone
Comments
Closed
garrgravarr
pushed a commit
that referenced
this issue
Jan 13, 2023
For MPD and MCS, the data rows are handled as vectors. MPD then calculates distances by calculating the Gower's distance matrix. MCS calculates the cosine between the two data vectors by dividing the inner product of the vectors, divided by the product of their magnitudes. The categorical entries are multiplied by comparing their values and if they aren't of the same value, the result is 0, otherwise 1. Similar for distances. For two elements of the same value, their difference is 0, otherwise 1. The above can only be calculated for both values being represented by a number. Entries, in which one or both of the vectors have NaN or None, are dropped from both vectors. Issues: closes #4, closes #6 sdv_metric.py contains a wrapper function, which can be used to produce functions which can be hooked into the ASyH ScoringHook.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Adaptation of SDV model (number of hidden layers for encoding and decoding, etc.) is done at the SDV model's construction, via arguments.
Models have different keyword arguments to specify their internal layout, therefore, every specific ASyH model needs to construct an argument list from the input data layout.
For keyword arguments, a dictionary should be used.
The dictionary can then be used as argument as in
sdv_model_constructor(**argument_dict).The text was updated successfully, but these errors were encountered: