TI patent publications disclose machine-learning methods for circuit verification and debug
Two U.S. patent applications assigned to Texas Instruments disclose complementary machine-learning techniques for analog or hybrid analog/digital integrated circuits. US20230409789A1 describes learning from simulated signal values and comparing predicted outputs with actual IC outputs for validation; US20230409790A1 describes inverse models that infer inaccessible internal signals or state-machine states from top-level inputs and outputs for debug. The documents also describe simulation-test selection using coverage metrics. These are patent disclosures, not evidence of product implementation or a deployed verification service.
Evidence
US20230409789A1 — Machine learning techniques for circuit design verification ↗
The patent record identifies the December 21, 2023 publication, Texas Instruments assignee, inventors, and the disclosed ML prediction and circuit-validation method.
US20230409790A1 — Machine learning techniques for circuit design debugging ↗
The related patent record identifies the same publication date and TI assignee and describes inverse ML models for predicting internal signals and states for circuit debug.