Non-Deep Active Learning for Deep Neural Networks

. 2022 Jul 13;22(14):5244. doi: 10.3390/s22145244

Algorithm 1 Sampling from Uncertainty Indicator

Input:

D_{L}

: labeled data pool,

D_{U}

: unlabeled pool, T: random initialized task model, C: number of cycles in active learning, N: number of samplings;
Output: Final learned parameters of T;

1:
for $i = 1$ to C do
2:
train T with $D_{L}$
3:
for $x_{i}$ in $D_{U}$ do
4:
$V = T (x_{i})$
5:
$i n d i c a t o r_{i} = U I G (V)$
6:
if ( $i n d i c a t o r_{i}$ is top-N) then
7:
$y_{i} = O R A C L E (x_{i})$
8:
$D_{L} = D_{L} + (x_{i}, y_{i})$
9:
$D_{U} = D_{U} - x_{i}$
10:
end if
11:
end for
12:
end for