Interpolation in a general sense is the process of estimating unknown values that fall between known values. In image processing, interpolation often refers to resizing images – e.g., computing pixel values when scaling or rotating an image (common methods: nearest neighbor, bilinear, bicubic interpolation). In the context of model training, one might talk about interpolation vs extrapolation: interpolation means making predictions on inputs that lie within the range of the training data (which is generally safer and more reliable), whereas extrapolation is trying to predict outside that range (which is harder and often yields larger errors). In some contexts, interpolation can mean constructing new data points within the known data points (like interpolating a function). In deep learning latent spaces, interpolation might refer to taking two latent vectors and linearly interpolating between them, then decoding to see a smooth transition in the output (often used to demonstrate a model has learned a continuous representation, e.g., in GAN latent space).