Machine learning inference queries are a type of database query for databases where a model pipeline is needed to evaluate its boolean predicates. Using a model zoo it is possible to select a variety of models to execute in a sequence rather than using a highly specialized model to answer every query predicate. Machine learning models can have multiple measurements for gauging performance however, and the quality of a query plan therefore is not only dependent on the time needed to compute it. Selecting a query plan of models that balances multiple objectives is not a trivial feat however. This work builds upon existing methods that utilize MIPs for model selection and ordering for machine learning inference queries by extending them with multi-objective optimizing capabilities. The opportunity for adding a third objective, namely memory footprint, to that of accuracy and execution cost is explored. Several methods are then considered and compared on their suitability, and the final chosen method, the Archimedean goal method, can generate Pareto optimal query plans that provide gains over naive, greedy methods. In addition, several methods of cutting down runtime on the original optimizer are explored, leading to a program than can generate higher quality solutions in less time.

Machine learning has been a computer sciences buzzword for years. The technology has a lot of potential and a huge number of applications that spoke to people with and without knowledge of computer sciences. Image, text and speech recognition, social profiling, computergames, everything seemed possible. Machine learning is not as much in the spotlight now since the rise of blockchain technology, but the field is still relevant and used widely. The field concerns itself with ‘teaching’ computers to make choices or recognize data by repeating experiments or looking at data, and then making changes to itself to achieve better results. Now consider the field of optimization. It is not as well known outside of the mathematics community but incredibly important nonetheless. It has a miriad of uses: flight scheduling, ambulance placement and route planning are a minimal selection of all the problems that Optimization gets used in. Optimization is about solving problems as quickly and optimally as possible, looking at ways to speed up the process or finding the solution with the lowest cost. Both fields have an extensive range of methods and applications but have not been combined a lot before. This report takes a versatile topic within machine learning, namely deep learning and combines it with a classic problem from Optimization. This model was first proposed in 2017, meaning that is a very young line of research. This report will further explore this crossover and look into the performance and possible applications. There will first be an introduction to both fields to make the report understandable for students from either mathematics or computer sciences faculties. The model will then be established and explained in detail. There will then be time devoted to looking into applications of the model and seeing how they perform in experiments. Lastly the results of the experiments will be discussed and their implications.