Machine Learning and Systems Programming, UC San Diego
Transformer-based models achieve strong performance for jet tagging at the CERN LHC, but deploying them in low-latency, resource-constrained trigger systems is challenging. We present an initial implementation of a quantized, integer-only transformer for jet tagging on the AMD Versal AI Engine (AIE). This framework provides a foundation for future research and is released as open-source software at github.com/KastnerRG/particle_transformer_aie.
Documentation for programmers interested in developing kernels and graph data-flow optimizations for the AMD AI Engine NPU. Includes explanations of key AI Engine concepts and code samples for common machine learning kernels.
Using Lagrangian mechanics to build the N-mass rigid pendulum system and utilizing 4th order Runge-Kutta method for numerical integration.
Generating local Lyapunov exponents, a measure of predictability given specific initial conditions, via a repeated process of Gram-Schmidt orthogonalization. The exponents are then plotted as a heatmap over the initial conditions of the system.
