deodorant.scaling-functions
Scaling functions for Deodorant.
->scale-details-obj
(->scale-details-obj theta-min theta-max log-Z-min log-Z-max)
Positional factory function for class deodorant.scaling_functions.scale-details-obj.
->scaling-funcs-obj
(->scaling-funcs-obj theta-scaler theta-unscaler theta-unscaler-no-centering log-Z-scaler log-Z-unscaler log-Z-unscaler-no-centering)
Positional factory function for class deodorant.scaling_functions.scaling-funcs-obj.
flatten-unflatten
(flatten-unflatten x)
Returns functions for flattening and unflattening the thetas. For example
when sampling from a multivariate normal theta will be a nested vector
TODO make me work for matrices
map->scale-details-obj
(map->scale-details-obj m__6522__auto__)
Factory function for class deodorant.scaling_functions.scale-details-obj, taking a map of keywords to field values.
map->scaling-funcs-obj
(map->scaling-funcs-obj m__6522__auto__)
Factory function for class deodorant.scaling_functions.scaling-funcs-obj, taking a map of keywords to field values.
scale
(scale data)
Normalizes data to lie inside a hypercube bounded at [-1 1] along.
Accepts a collection of data points [x] in which x may be a scalar or vector.
Returns a tuple [scaled-data unscale] containing the scaled data and
a function that inverts the transformation.
scale-points
(scale-points points)
Rescales points to a hypercube bounded at [-1 1].
Accepts a collection of points [x y] in which x is a D-dimensionl
vector and y is a scalar.
Returns a tuple [x-scaled y-scaled u5nscale-x unscale-y] containing
the scaled data and functions to revert the scaling.
setup-scaling-funcs
(setup-scaling-funcs scale-details)
Given a scale-details-obj returns a scaling-funcs-obj
unflatten-from-sizes
(unflatten-from-sizes sizes x)
update-scale-details
(update-scale-details scale-details scaling-funcs theta-new log-Z-new)