Inverse design and AI/Deep generative networks in food design: A comprehensive review

Background

Food material science has evolved to support the development of food products by connecting food structure, sensory, nutrition, food processing, and digestion with impact in consumers. However, food design has not evolved to deal with this increased complexity of food systems. And the ability to understand, capture the attention, and transform consumer demands into the chemical and physical attributes of the final product remains one of the biggest challenges in the food industry. As a result, new ways to support food design are necessary.

Scope and approach

This review describes the state-of-the-art of applications in food design utilizing artificial intelligence (AI)/Deep Generative Networks, including available resources and emerging capabilities and its relationship with the concept of inverse design.

Key findings and conclusions

Food design and formulation involve complex processes and many design parameters need to be considered while developing data-driven approaches. Most approaches identified are based on the association among ingredients, but less focus has been given to functional properties. Representation of data remains a real challenge and a very important research gap toward achieving a real and applicable concept of digital food design. Overall methods based on deep learning and natural language processing are the most utilized. Deep generative-based approaches have been rarely described and remain a critical research area.

NZBIDA