Research Data and Code for "Auditing Domain Alignment in Latent Representations for Cross-Domain Materials Machine Learning"

Key Points

• The aim is to establish a framework for auditing domain alignment in latent representations derived from diverse materials datasets.
• Implemented unsupervised latent encoder training for diverse datasets.
• Conducted representation-level validation to assess alignment quality.
• Utilized quantitative alignment diagnostics independent of predictive accuracy.
• Established framework to audit domain alignment effectively.
• Demonstrated that latent representations from different domains can be validated using the proposed methods.
• Provided tools for improving the reliability of machine learning models in materials science.