Representing Latent Dimensions Using Compressed Number Lines

Representation learning, Embedding, Neural network

Proposes compressed number lines as a method for representing latent dimensions—connecting ideas from cognitive science (mental number lines) to neural network representations.