Reference Issues/PRs

Fixes #1713
Fixes #578
Fixes #1057

What does this implement/fix? Explain your changes.

Simple fix to allow for correct parameter setting with GridCV. It removes a temporary measure to allow KNN to pass tests and fixes thanks to the new distance factory. Note we should do some work on making the valid parameters clearer for the different distance measures

Below code all works

code

knn = KNeighborsTimeSeriesClassifier(distance="dtw", n_neighbors=1, distance_params={
"window":0.5})
knn2 = KNeighborsTimeSeriesClassifier(distance="dtw", n_neighbors=1)

param_matrix = {"distance_params": [{"window": x/10.0} for x in range(0,10)]
}

CV Parameter search

grid = GridSearchCV(knn2, param_grid=param_matrix, cv=2, scoring="accuracy")
grid.fit(x_train, y_train)
print(pd.DataFrame(grid.cv_results_))
print(grid)
print(" params after CV = ",knn2.get_params())
knn4 = KNeighborsTimeSeriesClassifier(distance="dtw", n_neighbors=1, distance_params={
"window":0.0})

estimators = [
("knn", knn, [0, 1, 2]),
("knn2", knn2, [3, 4]),
("knn4", knn4, [5]),
]
x_train, y_train = load_basic_motions(split="train")
x_test, y_test = load_basic_motions(split="test")

train column ensemble

col_ens = ColumnEnsembleClassifier(estimators=estimators)
col_ens.fit(x_train, y_train)
print(" Score = ",col_ens.score(x_train, y_train))