AI is radically changing society and the main driver behind new AI methods and systems is machine learning. Machine learning focuses on finding solutions for, or patterns in, new data by learning from relevant existing data. Thus, machine learning algorithms are often applied to large datasets and then they more or less autonomously find good solutions by finding relevant information or patterns hidden in the data.

The search for nearest neighbors is an emerging and increasingly vital component in data analysis tasks, for example using vector embedding databases. Typically, the search is the bottleneck in terms of efficiency. Approximate nearest neighbor (ANN) search methods are often employed to speed up the application. However, different methods for ANN search come with different biases that can be positive or negative for the downstream application. In this project, the bias of different ANN methods and its impact on different applications will be studied.

Our focus is on improving optimization algorithms in online decision-making. Using techniques from online algorithms for solving optimization problems, we can provide worst-case guarantees, but normal (averagecase) behavior may not be satisfactory. Using techniques from machine learning, we can often provide good behavior in practice, but guarantees are lacking, and in particular missing for situations not captured by the training data. We aim at combining the best features from these two areas.