From Knowledge Transmission to Knowledge Construction: A Step towards Human-Like Active Learning

. 2020 Aug 18;22(8):906. doi: 10.3390/e22080906

Algorithm 1 Pool-based active learning with Information Capacity strategy

1:
procedureInformationCapacity( $m_{train}$ , $m_{test}$ )
2:
Initialize a labeled training set L;
3:
Initialize an unlabeled training pool $U = m_{train} - L$ ;
4:
Initialize a learning behavior of learners B with regard to a set of parameters $α, β, a, b$ ;
5:
Train a group of learners on the labeled set L;
6:
Measure performance of the group of learners on the test set $m_{test}$ ;
7:
Initialize several rounds $n_{round}$ and several queried examples $| L_{S} |$ ;
8:
for $r o u n d \in n_{round}$ do
9:
Estimate the probabilities with regard to (1) based on the learning behavior B;
10:
Sort the unlabeled items in U according to (4) based on the probabilities from the step 9;
11:
Query the items $L_{S}$ with the smallest of the maximum capacity S in a $r o u n d$ ;
12:
$L \leftarrow L \cup L_{S}$ ;
13:
$U \leftarrow U \ L_{S}$ ;
14:
Retrain a group of learners on the labeled set L;
15:
Measure performance of the group of learners on the test set $m_{test}$ ;
16:
end for
17:
return The performance of the learners with the interpretation of their learning behavior.