Fig. 2From: Using transfer learning and dimensionality reduction techniques to improve generalisability of machine-learning predictions of mosquito ages from mid-infrared spectraA schematic representation of a deep learning models that uses mosquito spectra as input to predict mosquito age classes. A CNN—no dimensionality reduction is applied: standardised spectral features are fed as input through four different convolutional layers, followed by one fully connected layer, with the predicted age classes shown as the output layer. B MLP—dimensionality reduction is used: spectral features that have been reduced in dimension using PCA or t-SNE are fed as input through 6 fully connected layers, with the predicted age classes shown as the output layerBack to article page