Within the last decade, novel machine learning (ML) models have introduced alternative ways to efficiently tackle long-standing quantum mechanics problems in physics and chemistry. Potential energies have been fitted [1,2], laws of physics have been rediscovered , atomization energies of organic molecules can be predicted , and clusters identified  and compounds mapped . ML can also be used to discover new molecules  or crystals , and even new reactions . Various properties and systems have been studied with ML, including electrons , chemical potentials , ionic forces , and molecular properties . Efforts to further improve and assess ML models for their application throughout compositional space are ongoing . When it comes to the acceleration of well established QM methods applied to surface science, however, ML based investigations, such as Refs. [15, 16], are sparse. The goal of this workshop is to bring together ML and surface scientists to explore potential future joint research efforts and collaborations. Effectively, we hope to jump start a new ML based sub-community within the larger field of theoretical surface science.
The goal of this workshop is to bring together the main scientific players who are likely to deepen our understanding of interfaces using QM and ML. We have contacted representative researchers from diverse scientific areas who all have already been involved in this topic in one way or another. They will all strongly benefit from the mutual exchange of ideas regarding this very important topic. It therefore comes as no surprise that they all already agreed to participate if this workshop is funded. The specific scientific domains include statistical mechanics, liquid and solid state physics, quantum chemistry, graph theory, molecular physics, condensed matter physics, optimization algorithms, data mining, and statistical learning.
We find it best to formulate the specific ideas and topics, to be discussed and tackled, in the form of questions.
-How can the interface structures and processes be identified and searched most effectively using ML?
-How can ML be used to model interfaces?
-What is the respective role of unsupervised and unsupervised ML for interface properties and processes?
-How much QM training data is required?
-Is DFT a sufficiently accurate reference method for ML of interfaces?
-What are good representations and kernels to construct ML models of interface systems?
If you wish to attend, please send us, using the Apply form, a short (MAX 1/2 page) statement of your interests and the reasons you wish to participate, and a very brief CV. We will peruse applications and inform you if (in view of the limited availabilities of place) we can accept your request by the end of March 2018.