Decision tree-based anomaly detection for trigger systems in high energy physics

We present a novel algorithm for anomaly detection using a decision-tree based autoencoder. The performance is demonstrated by identifying a simulated exotic Higgs decay against a background similar to what would be seen at the Large Hadron Collider. The fwX platform is used to implement the autoencoder on low-level firmware. We demonstrate that the latency and resource usage of such an implementation is adequate for the strict requirements at first-level trigger systems at the ATLAS and CMS experiments.