Neural Network Ensembles for Sensor-Based Human Activity Recognition Within Smart Environments

. 2019 Dec 30;20(1):216. doi: 10.3390/s20010216

Algorithm 5. Conflict resolution approach 4
Input: ${\vec{x}}_{i}$ , base models M_r, M_s
Output: class $y_{i}$
1:	$A c c_{t r a i n}^{r}$ represents training performance for base model $M_{r}$
2:	$A c c_{t r a i n}^{s}$ represents training performance for base model $M_{s}$
3:	$i f p_{i}^{r, 1} \times m_{r} \times A c c_{t r a i n}^{r} > p_{i}^{s, 1} \times m_{s} \times A c c_{t r a i n}^{s}$
4:	Then $y_{i} = {\hat{k}}_{i}^{r}$
5:	Else $y_{i} = {\hat{k}}_{i}^{s}$