Binary Neural Networks for FPGAs

This project explores the feasibility of implementing Binary Neural Networks (BNNs) for real-time tau identification in the ATLAS Level-1 trigger, in the context of the upcoming High-Luminosity LHC upgrades. BNNs, which operate with binary weights and activations, drastically reduce power and resource usage while maintaining high performance. These models are evaluated against Convolutional Neural Network (CNN) baselines in terms of efficiency and physics performance, and will be implemented on FPGAs using hls4ml for hardware deployment as part of the ATLAS - Next Generation Triggers upgrade.