Method of controlling for undesired factors in machine learning models

摘要: A method of training and using a machine learning model that controls for consideration undesired factors which might otherwise be considered by the trained during its subsequent analyzes new data. For example, may neural network on set images to evaluate an insurance applicant based upon image or audio data as part underwriting process determine appropriate life health premium. The is probabilistically correlate aspect applicant's appearance with personal and/or health-related characteristic. Any factors, such age, sex, ethnicity, race, are identified exclusion. receives (e.g., “selfie”) applicant, without considering suggests premium only remaining desired factors.