Material Discovery

While the previous chapter focussed on the development of ML modelsModels for property, this chapter discusses how those surrogate modelsModels can be used for discoveing novel materials. Here, we discuss various algorithms that can be employed for materials discovery. Various ML techniques are discussed, including ML-based optimization, material selection charts, generative artificial intelligence (AI) approaches such as generative adversarial networks (GANs) and variational autoencoders (VAEs), and reinforcement learningReinforcement learning. ML-based optimization enables efficient exploration of the vast design space, facilitating the identification of materials with desired properties. Material selection charts provide a systematic approach for screening and selecting materials based on key properties. Generative AI methods, like GANs and VAEs, offer exciting prospects for generating novel materials with specified characteristics, expanding the traditional design boundaries. Reinforcement learningReinforcement learning, a subfield of ML, guides the selection of candidate materials through sequential decision-making processes. The integration of ML techniques in materials discovery holds great promise, revolutionizing the field by enabling faster and more efficient identification of materials with tailored properties. Future prospects include advancements in ML algorithms, enhanced computational power, increased data availability, integration with experimental techniques, and the fusion of ML with emerging technologies like quantum computing and advanced imaging. These developments will drive innovation, leading to the development of advanced materials with improved performance and functionality.