AI self-quality assurance using learning curves in feedback loops

Description

One application of artificial intelligence (AI) in materials is the acceleration of materials innovation, which is the mission of the Materials Genome Initiative. However, to decrease the cost and time-to-market, we must continuously assess the quality of models with new facts. AI quality assurance (QA) is fundamental to determine a model's inadequacy. QA must be part of any AI system to oversee its learning process and direct it toward bettering quality or alerting to a quality issue. 

This work can

• Measure the accuracy of AI systems.
• Explore the adequacy of an AI models with regards to some data.
• Generally perform QA on AI systems’ learning process.
   

Major Accomplishments

• Gathering data sets and AI models
• Capturing the training and validation of AI models on data sets
• Generating learning curves from the training and validation steps

Additional Technical Details

Ongoing Challenges

• Gathering more datasets and AI models
• Use of Cloud of Reproducible Records (CoRR) to run the models and produce learning curves
• Design and implement the Python version of the self-QA framework
   

Associated Product(s)

Materials Genome Initiative