z = 1 ∑z =1 where z represents the latent vector of each deposit-related subcrop. This mean vector serves as a reference representation of the porphyry signature in latent space (See Figure 6). Figure 18 – Three-dimensional visualization of the latent space learned by the VAE-CNN model, showing the distribution of training data, latent representations of porphyry-related subcrops (reference latents), and their mean latent vector. 5.2 Similarity Estimation The similarity between the reference vector and all subcrops in the study area was quantified using the Euclidean distance: =∥ z −z ∥ where represents the distance between each subcrop and the porphyry signature. Since smaller distances indicate higher similarity, the distance values were transformed into a normalized similarity index ranging from 0 to 1: =1− − − 5.3 Prospectivity Map Generation The similarity values were spatially reassigned to their corresponding locations using a sliding window framework. To improve computational efficiency, predictions were performed using a stride of 10 pixels. A linear interpolation was subsequently applied to estimate similarity values at intermediate pixel locations. This process enabled the reconstruction of a smooth, highresolution prospectivity map across the study area (See Figure 7).
RkJQdWJsaXNoZXIy MTM0Mzk2