config.gin

import divik.cluster
import divik.core.gin_sklearn_configurables
import divik.feature_selection

# Macros:
# ==============================================================================
clustering = @DiviK()
feature_extraction = @NoSelector()

# Parameters for clustering/DiviK:
# ==============================================================================
clustering/DiviK.distance = 'correlation'
clustering/DiviK.fast_kmeans = @GAPSearch()
clustering/DiviK.features_percentage = 1.0
clustering/DiviK.filter_type = 'gmm'
clustering/DiviK.kmeans = @DunnSearch()
clustering/DiviK.minimal_features_percentage = 0.01
clustering/DiviK.minimal_size = 200
clustering/DiviK.n_jobs = -1
clustering/DiviK.normalize_rows = True
clustering/DiviK.rejection_percentage = None
clustering/DiviK.rejection_size = 2
clustering/DiviK.use_logfilters = True
clustering/DiviK.verbose = True

# Parameters for clustering/DunnSearch:
# ==============================================================================
clustering/DunnSearch.drop_unfit = True
clustering/DunnSearch.inter = 'closest'
clustering/DunnSearch.intra = 'furthest'
clustering/DunnSearch.kmeans = @divik.cluster._kmeans._core.KMeans()
clustering/DunnSearch.max_clusters = 10
clustering/DunnSearch.method = 'auto'
clustering/DunnSearch.min_clusters = 2
clustering/DunnSearch.n_jobs = -1
clustering/DunnSearch.n_trials = 10
clustering/DunnSearch.sample_size = 1000
clustering/DunnSearch.seed = 42
clustering/DunnSearch.verbose = True

# Parameters for experiment:
# ==============================================================================
experiment.exist_ok = True
experiment.model = @Pipeline()
experiment.omit_datetime = True
experiment.steps_that_require_xy = None
experiment.verbose = True

# Parameters for clustering/GAPSearch:
# ==============================================================================
clustering/GAPSearch.drop_unfit = True
clustering/GAPSearch.kmeans = @divik.cluster._kmeans._core.KMeans()
clustering/GAPSearch.max_clusters = 2
clustering/GAPSearch.min_clusters = 1
clustering/GAPSearch.n_jobs = -1
clustering/GAPSearch.n_trials = 10
clustering/GAPSearch.sample_size = 1000
clustering/GAPSearch.seed = 42
clustering/GAPSearch.verbose = True

# Parameters for clustering/_core.KMeans:
# ==============================================================================
clustering/_core.KMeans.distance = 'correlation'
clustering/_core.KMeans.init = 'kdtree_percentile'
clustering/_core.KMeans.leaf_size = 0.01
clustering/_core.KMeans.max_iter = 100
clustering/_core.KMeans.n_clusters = 1
clustering/_core.KMeans.normalize_rows = True
clustering/_core.KMeans.percentile = 99.0

# Parameters for feature_extraction/NoSelector:
# ==============================================================================
# None.

# Parameters for Pipeline:
# ==============================================================================
Pipeline.memory = None
Pipeline.steps = \
    [('feature_extraction', %feature_extraction), ('clustering', %clustering)]
Pipeline.verbose = False