Example with GaussianMixtureClusterer about betaml.jl HOT 3 CLOSED

Thank you. I have read your message, I'll have a look later this evening or tomorrow.

from betaml.jl.

Hello, thank you for reporting.
A bug slipped in the code evading the unit tests during the reorganisation of the package. It should now be ok in the v0.9 that I am going to release in minutes, so please update your packages to use BetaML 0.9.

Note that in the MLJ interface the name of the model is now GaussianMixtureClusterer and the number of classes parameter is n_classes.

Sorry for the confusion, but it was important to reorganise the package API. It should be much more stable now. You can read all the model names here.

Here you have the specific model hyperparameters:

• here and here for the BetaML GMMClusterer model
• here for GaussianMixtureClusterer (MLJ interface)

An example of usage would then be the following:

import MLJBase
const Mlj = MLJBase
using BetaML

X, y                        = Mlj.@load_iris
model                       = GaussianMixtureClusterer(mixtures=FullGaussian,n_classes=3)
modelMachine                = Mlj.machine(model, X) # DimensionMismatch
(fitResults, cache, report) = Mlj.fit(model, 0, X)
yhat                        = convert(Vector{Int64},Mlj.predict_mode(model, fitResults, X))
ytrue                       = fit!(OrdinalEncoder(),y)
accuracy(ytrue,yhat,ignorelabels=true) # 0.987

Let me know if using BetaML v0.9 is it now working on your side !

from betaml.jl.

Thank you @sylvaticus , problem solved 👍🏽

from betaml.jl.