Driving Moore's Law with Python-Powered Machine Learning: An Insider's Perspective
summary
Claims that machine learning keeps Moore’s Law moving by absorbing manufacturing variation into models and CAD tools, with concrete stacks: Python 2.7 with numpy scipy scikit-learn, SPICE simulators and C kernels for speed. Methods include genetic programming, elastic‑net / FFX regression, Gaussian‑process surrogate models and Mer‑Twister RNG. Results: white‑box models within ~10% error, rare‑event estimation cuts two months on 100 cores to 20 minutes on 10 cores, and corner analysis solves 3,000 corners in 275 simulations.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 1:14Moore's Law across industries
- 3:42Moore's Law challenges
- 7:10Variation in chip design
- 12:05ML for variation abstraction
- 13:50Python symbolic modeling
- 21:56Scaling white-box models
- 27:12Power consumption estimation
- 31:24Rare event estimation
- 36:38Genetic programming topologies
- 39:06Corner analysis optimization
- 42:39Sampling techniques
- 43:22Conclusion
related talks
43mKeeping Moore's law alive: Neuromorphic computing
01h 53mMachine Learning for Manycore System Design and Optimization
26mKeynote: Artificial Innovation Accelerates Artificial Intelligence
58mThe Next Decade of Software Development
54mProgramming Language Design and Implementation for the Machine Learning Era: A Personal(…)
33mMachine Learning explained to humans
01h 10mBayesian Optimization and Machine Learning for Accelerating Experiments in the Physical Sciences
30mCultivating a company-wide machine learning culture