Algorithm 2 DWF method
(2-1) The training and fitting stages.
Require: Training datasets, St = {(x1, y1), …, (xmt, ymt)}, t = 1, …, T, and the mean measurement time, wt, of each dataset.
1:	for t = 1, …, T do
2:	Train a classifier, ft, on St;
3:	end for
4:	for t = 1, …, T do
5:	Estimate the optimal weights, { ${\beta }_1^t$,… ${\beta }_T^t$}, of {f1, …, fT} for St using the appropriate technique;
6:	end for
7:	Receive a T × T matrix, ${\beta }_i^j$;
8:	For a classifier, i.e., ft0, fit curve, Ct0(w), with {( ${\beta }_{t_0}^1$, ω1), …, ( ${\beta }_{t_0}^T$, ωT)};
9:	Revise the t0th row vector, βt0, as [Ct0(w1), …, Ct0(wT)], by means of scaling each column vector of ${\beta }_i^j$;
10:	for t = 1, …, T except t0 do
11:	Fit curve Ct(w) with {( ${\beta }_t^1$, w1), …, ( ${\beta }_t^T$, wT)};
12:	end for
Ensure: The classifiers, ft, and the corresponding fitting functions, Ct(w), t = 1, …, T.
(2-2) The testing stage.
Require: Test the dataset, XT+n, and the mean measurement time, wT+n, n > 0; the classifiers, ft, and the corresponding fitting functions, Ct(w), t = 1, …, T.
1:	for t = 1, …, T do
2:	Calculate the weight of ft at time wT+n, namely Ct(wT+n);
3:	end for
4:	Test XT+n using the classifier ensemble, $h_{T+n}={\sum }_{t=1}^T{C}_t\left(w_{T+n}\right)f_t$;
5:	Estimate the labels: YT+n = hT+n(XT+n);
Ensure: Estimated labels YT+n.