Extended Data Fig. 2: Architecture and details of Residual Artificial Neural Network (RANN).
From: Deploying photovoltaic systems in global open-pit mines for a clean energy transition
The input of the RANN is a 1 × 16 vector, which consists of five physical geographical factors, six socio - economic factors, and five resources condition factors (as shown in the leftmost box with a yellow background). After passing through 50 residual blocks of size 1 × 1024 in the middle, it finally outputs a 1 × 1 vector, which represents the probability. Specifically, these 50 residual blocks in the middle use shortcut connections to avoid the information loss problem in the deep network (as shown in the middle box with a gray background). The output 1 × 1 vector is the probability of PV deployment for each mining patch.
