Paragraph Number

180

Learning can utilize one or more cost functions, e.g. the cost being an excellent or a poor clinical outcome. The cost function can yield information of how far a particular solution, e.g. a clinical treatment, treatment sequence or treatment algorithm or surgical technique, is from an optimal outcome, e.g. an excellent score in a patient reported outcome measure. ANNs can find the solution, e.g. a clinical treatment, treatment sequence or treatment algorithm or surgical technique, that yields the lowest cost, e.g. distance or amount away from an optimal outcome or excellent score in a patient reported outcome. The cost can be a function of the observations. The cost can be described as a statistic. A cost can be the mean squared error, which can try to minimize the average squared error between the network's output and one or more target values over example pair(s). A cost function can be selected or predetermined for a particular problem set, e.g. a clinical problem set or clinical observation data, e.g. pre-operative, intra-operative or post-operative data. AI can find and develop one or more optimal cost functions for a set of observational data and AI can refine the cost function as the size of the observational data set increases. (See US Patent 11,278,413 Para. 58-63).

Added by DJM Jan 2024