CMS is a general-purpose detector at LHC. During a run, it generates about 40 TB data per second. Since It is not feasible to readout and store such a vast amount of data, so it selects and stores only interesting events or events likely to reveal new physics phenomena. Currently, Boosted Decision Trees (BDTs) with handcrafted features are used for that.
The goal of the project is to test the performance of different deep learning algorithms and minimize the latency in inference. The best models in each algorithmic type will be wrapped into a functional prototype emulator. Other than this, a significant emphasis will be given on proper documentation.