Open Access

Table 2

Performed experiments. From left to right: experiment code, dimensionality of the UMAP projection of the autoencoded embeddings (see text), DBSCAN ϵ value, Calinski-Harabasz index, Davies-Bouldin index, number of clusters, average cluster size, and number of outliers.

Experiment #UMAP dims ϵ CHI DBI #Cluster < Size > #Outlier
A1 4 0.30 4.15 2.27 4 7.0 150
A2 4 0.35 4.58 2.30 9 7.0 115
A3 4 0.40 6.58 2.59 7 14.3 78
A4 4 0.45 8.79 2.98 5 24.8 54

B1 6 0.30 5.03 2.06 2 7.5 163
B2 6 0.35 4.70 2.50 10 7.8 100
B3 6 0.40 6.39 2.65 8 11.9 83
B4 6 0.45 7.79 2.84 7 17.4 56

C1 8 0.30 4.99 2.10 3 7.3 156
C2 8 0.35 5.00 2.36 8 7.9 115
C3 8 0.40 5.95 2.73 8 11.5 86
C4 8 0.45 12.2 3.28 3 36.3 69

D1 10 0.30 4.68 2.20 3 8.0 154
D2 10 0.35 4.66 2.43 8 8.0 114
D3 10 0.40 7.06 2.83 7 14.4 77
D4 10 0.45 7.58 2.83 7 18.9 46

E1 12 0.30 5.54 2.11 3 8.0 154
E2 12 0.35 5.16 2.51 6 8.5 127
E3 12 0.40 6.29 2.65 8 11.8 84
E4 12 0.45 8.24 2.86 6 20.0 58

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.