Periodic Reporting for period 1 - hls4ml (High Level Synthesis for Machine Learning)
Période du rapport: 2021-04-01 au 2022-09-30
The hls4ml project consisted in the development of an open-software library to process pre-trained neural networks and convert them to an electronic circuit. One could then print these circuits on custom AI chips or execute them on Field-programmable gate arrays, universal circuit-emulator chips. The hls4ml library could be used to develop edge-computing AI devices, such as wearable monitoring devices for medical applications, real-time data processing on custom hardware, etc. The library was designed having in mind the real-time process of high-energy proton-proton collisions at the CERN Large Hadron Collider. Its design makes it suitable for fast-inference applications, e.g. high-rate trading.