Fig. 2: The latent spaces of two-way, three-way, and four-way oil mixtures. | Nature Communications

Fig. 2: The latent spaces of two-way, three-way, and four-way oil mixtures.

From: Pattern recognition based on machine learning identifies oil adulteration and edible oil mixtures

Fig. 2

GMM simulated pure oil samples were mixed by linear combination of ratios drawn from a Dirichlet distribution and visualized in the latent space. Gradient legend (blue to red) used for (ac), denoting the purity of oil measured against the major oil. Ten colors were used (df) to represent relative proportion of mixtures in the pie charts. a, b Ten epicenters in the two-way mixtures describe the ten edible oils in the purest form (red). The purity shifts to 50% (blue) as two oils are mixed in equal proportions, and then shifting back to red, in the epicenter of another pure oil type as indicated by the pie charts. c, d Three-way and e, f four-way mixtures exhibit the same patterns, even though the epicenters become fuzzier due to the complexity of mixtures.

Back to article page