Algorithm 2.

The ResNet with MMM for indoor positioning.

Input: RSSI datasets R
Output: ResNet model
1	Set the initial value of timewindow and hyperparameters (see Table 1)
2	forT$\leftarrow$start to end in a step of timewindowdo
3	for each t$\in$Tagsdo
4	for$\tau$$\leftarrow$Tk to Tk + 1 in a step of mini-spando
5	for each g$\in$Gatewaysdo
6	$Ig\leftarrow \left[Min\left(Rt\right);Mean\left(Rt\right);Max\left(Rt\right)\right]$
7	end for
8	Pn$\leftarrow$ [I1, …, Ig]
9	end for
10	Form input matrix across time IM$\leftarrow$ [P1, …, Pn] with a label l
11	end for
12	end for
13	Randomly split matrices for training, validation and testing
14	Construct ResNet structure (three blocks of skip connection)
15	for epochs with a batch size (m = 10):
16	$\phi \leftarrow 1/m·{\nabla }_{\theta }\sum _{\mathrm{i}}\text{Loss}\left(F\left(\left[x_1,x_2,\dots ,x_n\right],\theta \right),O_{E,i}\right)$
17	Adam algorithm (Same as Algorithm 1)
18	end for
19	return ResNet model