This file can create a semi-supervised data set in one of two ways.
-
The standard way. We iterate through the code at hand and we predict on unlabled data, some label. If this label is within some confidence interval, we update the label and append it to the data. We continue iteration til we include no new data.
-
The new way. We iterate through the code at hand and predict on unlabled data. Then as we add to the unlabled data, our features include some statistics That describe the data up to some time. We now update those statistics at hand. This is more sophisticated approach found in lines 288- 289
#runner.py
This file can be used to iterate the semi-supervised model n times. Often we will need to create statistically significant samples for us over training and test data to get some sense of how a method performs. This code will iterate through the semi supervised module n times and output some result.