Using Machine Learning to Optimize Semiconductor Test

Machine learning (ML) is a field of engineering under the artificial intelligence (AI) umbrella which is based on the idea that systems can learn from data, identify patterns and make decisions with minimal human intervention. ML is often used as a prediction tool to perform analysis on large amounts of data, automate analytical model building, and provide the user (human, machine, or calling function) with event occurrence probability. To date, the test community has not embraced ML and continues looking at ML with skepticism when approached with the paradigm shift - replacing conventional product test suites with a test suite selected by an ML algorithm. The skepticism could be the result of a combination of multiple factors such as limited familiarity with ML, partial awareness of its capabilities, lack of tools that can be easily used, and the workload involved in the process of training a machine to discern data. Additionally, by running a compressed test suite suggested by ML, questions arise about test escapees and the potential impact on product quality. This talk uses the example of an Open-Short test to throw light on optimizing semiconductor test using ML. Presenter: Gerard John, Sr. Director – FCBGA Presented at EDPS 2020