Disentangled representations is an unsupervised learning technique that breaks down, or disentangles, each feature into narrowly defined variables and encodes them as separate dimensions. The goal is to mimic the quick intuition process of a human, using both “high” and “low” dimension reasoning