Calorimeter Image Convolutional Anomaly Detection Algorithm (CICADA) uses low level Compact Muon Solenoid experiment's trigger calorimeter information as the input to convolutional autoencoder to find anomalies produced during Large Hadron Collider proton-proton collisionsis. Quantization Aware Training and Knowledge Distillation are used to compress the model for sub-500ns inference on Field-Programmable Gate Arrays.

The publicly available repo can be found here, and the CICADA website is here.

Fast Inference for Rare Events based on Features in Liquid-argon ionization imagerY (FIREFLY) uses raw waveform data from Liquid Argon Time Projection Chamber experiments as the input to convolutional autoencoders to detect anomalous interactions or features in the data. Inspired by CICADA, we use similar techniques to compress models for microsecond-scale inference on Field-Programmable Gate Arrays while our input images can be significantly larger.

A publicly available repo with a tutorial on how a model can be optimized for FPGA deployment can be found here.