A TI patent publication describes machine-learning-assisted generation of real-number models

U.S. patent application US20250117560A1, assigned to Texas Instruments, describes obtaining analog-simulation logs, training a neural network, deriving a mathematical transfer function, and producing a SystemVerilog-compatible simulation kernel for real-number modeling. This Event records a published patent disclosure; publication does not establish product implementation, internal deployment, or the scope of any deployed flow.