Figure. 5 | Scientific Reports

Figure. 5

From: Aiding food security and sustainability efforts through graph neural network-based consumer food ingredient detection and substitution

Figure. 5

(a) Example of the ingredients prediction input graph. Nodes contain binary encoded ingredients vectors, and undirected links exist between products that are made by the same brand. This task involves the binary classification of each node, which equates to the model’s prediction of whether the product contains the specified ingredient. (b) Example of the food substitution graph. Nodes contain binary encoded ingredients vectors, and directed links exist from a product to be replaced, to a viable substitution. The model must predict the existence of these links, compared to the existence of negatively (random) links that do not satisfy the substitution criteria.

Back to article page