Genetic Algorithm (GA)

A combinatorial search methodology that is based on natural genetics.  In Genetic Algorithms, alternative solutions are evaluated and ranked by performance and the attributes of higher performers are combined through cross-over and mutations to create new solutions.  By repeating this process, new solutions become more fit.  GAs deliver near-optimal solutions quickly by sampling a relatively small number of alternatives from an often massive search space.  In contrast to Gradient Descent/Ascent, GAs can search highly complex discontinuous functions with many local minima/maxima with good results.

Gradient Descent / Ascent

A search methodology that efficiently descends or climbs a continuously smooth curve or surface seeking the optimal (global minimum or maximum).  In contrast to Genetic Algorithms, Gradient Descent/Ascent does not work well with highly complex (many peaks and valleys) or discontinuous functions because it may get "stuck" in local minima/maxima or at discontinuity boundaries.

In-Sample Data

Data that is used for modeling or pre-processing optimization of data prior to modeling.  Performance on this data is not suitable for making independent decisions about the model because it was used to create the model and will look to perform better than in reality going forward.  In-Sample data is in contrast to Out-of-Sample data.