AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Machine Learning Models Generalize Fire Interpolation
Machine learning models generalize fire interpolation, or that computer vision data sets are interplative. If you take any two images and you interpolate between them, what do you get? You just get a faded copy of both. A new paper shows that the models are not interprlative in their latent space either. The probability of test data being in the convex hull of your training data is near zero past a certain number of dimensions. Randall Bell rerio says we need to have an entirely new definition of interpolation.