The DPA package is the scikit-learn compatible implementation of the Density Peaks Advanced clustering algorithm. The algorithm provides robust and visual information about the clusters, their statistical reliability and their hierarchical organization.
BSD 3-Clause "New" or "Revised" License
27
stars
9
forks
source link
Automatically set initial parameters for consecutive runs of the DPA estimator #3
When running the DPA clustering for different values of the parameter Z, all attributes are computed each time, i.e. indices and distances of nn-neighbors, which makes the analysis computationally inefficient.
The algorithm performance would improve if consecutive runs could use the computed attributes as values for the corresponding __init__ parameters, and change only Z, as the parameter of interest.
The following example shows how to manually set each __init__ parameter value to the corresponding attribute computed in the latest run of the algorithm, by using the scikit-learn set_params function:
# Initialization and first run
est = DPA.DensityPeakAdvanced(Z=1.5)
est.fit(data)
# Set each __init__ parameter value to the new computed value
est.set_params(nn_indices=est.nn_indices_)
# Run set_params for all the computed parameters
[...]
# Set the new value of the parameter of interest Z
est.set_params(Z=1)
# Run the DPA clustering again
est.fit(data)
It would be useful to set the values of those __init__ parameters automatically, so that only the parameter Z has to be changed.
When running the DPA clustering for different values of the parameter Z, all attributes are computed each time, i.e. indices and distances of nn-neighbors, which makes the analysis computationally inefficient. The algorithm performance would improve if consecutive runs could use the computed attributes as values for the corresponding __init__ parameters, and change only Z, as the parameter of interest.
The following example shows how to manually set each __init__ parameter value to the corresponding attribute computed in the latest run of the algorithm, by using the scikit-learn
set_params
function:It would be useful to set the values of those __init__ parameters automatically, so that only the parameter Z has to be changed.