Over the last few years we have seen an explosion in new techniques applying Machine Learning (ML) to improve traditional algorithms, data structures, or systems in general across all fields in computer science from databases to networks and theory. For example, there has been work on improving query optimization, indexing, storage-layouts, scheduling, log-structured merge trees, sorting, compression, sketches, among many other things using ML. Arguably, the motivation behind these techniques are similar: machine learning is used to capture something about the data and/or workload to derive a better algorithm or data structure. Ultimately, what these techniques will allow us to build are “instance-optimized” systems; systems that self-adjust to a given workload and data distribution to provide unprecedented performance and avoid the need for tuning by an administrator.

​

As this field is moving extremely fast and is scattered around several research areas, it is increasingly harder to keep up with and start working in this area. In order to make it easier to navigate all the rapidly appearing new techniques, we are trying something: a workshop tutorial. The idea is to invite researchers, who are active in the area, to present their already published work in a survey like fashion, provide an update on what they are currently working on, and offer their take on open research questions and biggest hurdles to make the techniques practical.