Opportunity at National Institute of Standards and Technology (NIST)
Tailored Machine Learning Oracles to Guide Data Processing for Statistical Inference
Information Technology Laboratory, Statistical Engineering Division
Please note: This Agency only participates in the February and August reviews.
Statisticians must often decide about whether/how to process collected data preparatory to making a statistical inference. For example, a decision may be needed about the choice of imputation technique to address missing data, or whether to modify data to reflect known constraints such as convexity or monotonicity. In such situations the statistician would like to be able to rapidly construct a machine learning (ML) oracle to recommend a course of action. Ideally, the ML oracle would be tailored both to the collected data and to the statistician's specific inferential goal. For example, the oracle might advise for or against using the pooled adjacent violator's algorithm to account for model monotonicity before estimating a model change-point. Or, in the case of missing data, the constructed oracle would recommend a choice of imputation technique.
The goal of this research opportunity is to identify settings in which a tailored oracle can be effective, and to study in those settings the process of oracle construction and the oracle's effectiveness for the planned inference. This work requires doctoral-level understanding of mathematical statistics. Also welcome would be some familiarity with machine learning theory and experience with a coding environment for developing artificial neural networks. Post-doctoral research in the Statistical Engineering Division of the National Institute of Standards and Technology also offers opportunities for statistical collaboration with new and ongoing NIST projects.
machine learning; artificial neural network; data preparation; statistical inference
Open to U.S. citizens
Open to Postdoctoral applicants