Method and apparatus for creating an extraction model using Bayesian inference

• US 6,883,148 B1
• Filed: 01/31/2002
• Issued: 04/19/2005
• Est. Priority Date: 08/28/2001
• Status: Expired due to Term

• Current Assignee: Cadence Design Systems Incorporated (Cadence Group)
• Original Assignee: Cadence Design Systems Incorporated (Cadence Group)
• Inventors: Teig, Steven, Chatterjee, Arindam
• Primary Examiner(s): Thompson, A. M.
• Assistant Examiner(s): Kik, Phallaka
• Application Number:
  US10/062,193
• Time in Patent Office: 1,174 Days
• Field of Search: 716/4, 703/2, 703/15, 706/920, 706/47
• US Class Current: 716/115
• CPC Class Codes: G06F 30/367 Design verification, e.g. using simulation, simulation program with integrated circuit emphasis [SPICE], direct methods or relaxation methods
• CPC Class Codes: Y10S 706/92 Simulation

Abstract

A system for using machine-learning to create a model for performing integrated circuit layout extraction is disclosed. The system of the present invention has two main phases: model creation and model application. The model creation phase comprises creating one or more extraction models using machine-learning techniques. First, a complex extraction problem is decomposed into smaller simpler extraction problems. Then, each smaller extraction problem is then analyzed to identify a set of physical parameters that fully define the smaller extraction problem. Next, models are created using machine learning techniques for all of the smaller simpler extraction problems. The machine learning is performed by first creating training data sets composed of the identified parameters from typical examples of the smaller extraction problem and the answers to those example extraction problems as solved using a highly accurate physics-based field solver. The system them uses the created training sets to train neural networks that will be used to model the extraction problems. Bayesian inference is used to train the neural networks models. Bayesian inference may be implemented with normal Monte Carlo techniques or Hybrid Monte Carlo techniques. After the creation of a set of models for each of the smaller simpler extraction problems, the machine-learning based models may be used for extraction.

Claims

1. A method of constructing a model for estimating at least one electrical characteristic for an extraction sub-problem, said method comprising:

identifying a set of physical measurements of integrated circuit components that define said extraction sub-problem;

selecting a set of training cases for said specific extraction sub-problem, each of said training cases including an associated set of said physical measurements;

solving said specific extraction sub-problem for each of said training cases using said associated set of physical measurements as an input to an accurate physics based model to generate an associated output; and

training a machine-learning model with Bayesian inference using said associated set of physical measurements and associated outputs as training data.

2. The method as claimed in claim 1 wherein said electrical characteristic comprises capacitance.

3. The method as claimed in claim 1 wherein said electrical characteristic comprises resistance.

4. The method as claimed in claim 1 wherein said extraction sub-problem comprises a section of interconnect wire and nearby interconnect wiring within a define halo.

5. The method as claimed in claim 1 wherein said extraction sub-problem comprises a section of interconnect wiring.

6. The method as claimed in claim 1 wherein one of said set of physical measurements comprises a spacing between a pair of interconnect lines.

7. The method as claimed in claim 1 wherein one of said set of physical measurements comprises a wire width.

8. The method as claimed in claim 1 wherein one of said set of physical measurements comprises a wire length.

9. The method as claimed in claim 1 wherein selecting a set of training cases comprises randomly generating input measurements with a gamma probability distribution.

10. The method as claimed in claim 1 wherein said electrical characteristic comprises delay.

11. The method as claimed in claim 1 wherein said machine-learning model comprises a neural network.

12. A computer-readable medium, said computer-readable medium comprising a set of instructions for constructing a model for estimating at least one electrical characteristic for an extraction sub-problem by performing the steps of:

identifying a set of physical measurements of integrated circuit components that define said extraction sub-problem;

selecting a set of training cases for said specific extraction sub-problem, each of said training cases including an associated set of said physical measurements;

solving said specific extraction sub-problem for each of said training cases using said associated set of physical measurements as an input to an accurate physics based model to generate an associated output; and

training a machine-learning model with Bayesian inference using said associated set of physical measurements and associated outputs as training data.

13. The computer-readable medium as claimed in claim 12 wherein said electrical characteristic comprises capacitance.

14. The computer-readable medium as claimed in claim 12 wherein said electrical characteristic comprises resistance.

15. The computer-readable medium as claimed in claim 12 wherein said extraction sub-problem comprises a section of interconnect wire and nearby interconnect wiring within a define halo.

16. The computer-readable medium as claimed in claim 12 wherein said extraction sub-problem comprises a section of interconnect wiring.

17. The computer-readable medium as claimed in claim 12 wherein one of said set of physical measurements comprises a spacing between a pair of interconnect lines.

18. The computer-readable medium as claimed in claim 12 wherein one of said set of physical measurements comprises a wire width.

19. The computer-readable medium as claimed in claim 12 wherein one of said set of physical measurements comprises a wire length.

20. The computer-readable medium as claimed in claim 12 wherein selecting a set of training cases comprises randomly generating input parameters with a gamma probability distribution.