Skip to content

specifying distance functions #108

@andrewheusser

Description

@andrewheusser

There is an inconsistency with how distance functions for different feature dimensions are handled. When an egg is created, the user can optionally pass a dist_funcs dictionary containing distance functions for each feature. This is used in the clustering analyses (fingerprint, temporal). However, when performing naturalistic analyses, a new distance function is assigned, e.g. egg.analyze('spc', match='best', distance='correlation'). This forces the distance function to be 'correlation' for all features. Some possible resolutions:

  1. keep the distance argument, but default to egg.dist_funcs. if a string or function is passed, assume that all features should use that distance function. if a dictionary of feature:function pairs is passed, use these instead of the functions specified in egg.dist_funcs

  2. get rid of the distance argument in the analyze method (always infer it from the egg.dist_funcs dict).

1 seems like a more flexible solution...by default assumes you want to use egg.dist_funcs, but allow the user to modify the distance functions on the fly. @jeremymanning and @paxtonfitzpatrick , can I get your thoughts on this?

Metadata

Metadata

Labels

No labels
No labels

Type

No type
No fields configured for issues without a type.

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions