Please enable JavaScript.
Coggle requires JavaScript to display documents.
LABELED/UNLABELED DATA - Coggle Diagram
LABELED/UNLABELED
DATA
Clusters/Folds
$$K_2$$
Data extraction
Unlabeled data
Active learning
extraction
fold random sampling
OR
fold weighted sampling
Oracle(s)
Add new labeled data
Labeled data
Test set
n
samples
per each fold
symmetric random
extraction
Train set
n residuals per
each fold
ML classification
training
Cross-cluster/fold
validation
Constraints
Distance between
train & test points
stratification criteria
$$K_1$$
$$K_n$$
Constraints
balance class distribution
same cluster size (aprox.)
balance labeled/unlabeled ratio
account for dead zones
min spatial autocorrelation
...