I work with an unsupervised training procedure that makes individual neurons in an associative memory represent either features, or exemplars from the training set. Finding the latter always makes me happy. When inspecting a recent run I found this lonely horse. ❤️ (Weights as images in a grid.)