Arindam M Chatterjee, 55Cupertino, CA

6687887, Feb 3, 2004

Jan 31, 2002

10/061437

Steven Teig - Menlo Park CA
Arindam Chatterjee - San Carlos CA

Cadence Design Systems, Inc. - San Jose CA

G06F 1750

716 5, 716 4

A system for using machine learning based upon Bayesian inference using a hybrid monte carlo method 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, complex mathematical 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 uses Bayesian inference implemented with a hybrid Monte Carlo method to train a set of neural networks for extraction problems.

6735748, May 11, 2004

Jan 31, 2002

10/062185

Steven Teig - Menlo Park CA
Arindam Chatterjee - San Carlos CA

Cadence Design Systems, Inc. - San Jose CA

G06F 945

716 5, 716 2, 716 4

A machine-learning model may be created to perform integrated circuit layout extraction. Using such a machine-learning system has two main phases: model creation and model application. The model creation phase comprises creating one or more extraction models using machine-learning techniques. 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. Next, the system performs machine learning using Bayesian inference in order to train the neural network models. The Bayesian inference may be implemented with normal Monte Carlo techniques, Hybrid Monte Carlo techniques, or other Bayesian learning 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.

6854101, Feb 8, 2005

May 9, 2003

10/434670

Steven Teig - Menlo Park CA, US
Arindam Chatterjee - San Carlos CA, US

Cadence Design Systems Inc. - San Jose CA

G06F017/50

716 5, 716 6, 716 10

The present invention introduces a method of quickly extracting the capacitance for interconnect wires in an integrated circuit routed with a non Manhattan architecture. To extract the capacitance a section containing non Manhattan wiring, the present invention proposes an approximation system that approximates the section of non Manhattan wiring with a Manhattan wiring section that has a capacitance per unit length that is generally proportional to the length of the approximated section. The capacitance effect from the approximated Manhattan wiring section may then be adjusted with a correction factor. Specifically, the present invention proposes that the capacitance be calculated for an interconnect wiring section by multiplying the length of the interconnect wiring section by an approximated capacitance per unit length value of a similar Manhattan wiring segment and adding a correction factor that corrects for the difference between the approximated Manhattan wiring section and the original non Manhattan wiring section.

6857112, Feb 15, 2005

Jan 31, 2002

10/062264

Steven Teig - Menlo Park CA, US
Arindam Chatterjee - San Carlos CA, US

Cadence Design Systems, Inc. - San Jose CA

G06F017/50

716 5, 716 4, 716 6

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 training sets are then used to train the models. In one embodiment, neural networks are used to model the extraction problems.

6880138, Apr 12, 2005

Dec 31, 2002

10/335097

Steven Teig - Menlo Park CA, US
Arindam Chatterjee - San Carlos CA, US

Cadence Design Systems, Inc. - San Jose CA

G06F017/50

716 4, 716 5, 716 6, 716 1

The present invention introduces novel methods of generating input vectors for machine learning system that will perform extraction. Experimental design is employed to select a set of training points that provide the best information. In one embodiment, a set of input vectors and output vectors are analyzed to determine a set of critical input parameters. Then, a spanning point generation program is used to generate a set of spanning points that cover the identified critical input space. The training point set then used to train a machine learning model.

6883148, Apr 19, 2005

Jan 31, 2002

10/062193

Steven Teig - Menlo Park CA, US
Arindam Chatterjee - San Carlos CA, US

Cadence Design Systems, Inc. - San Jose CA

G06F017/50
G06F017/11
G06F017/17
G06F017/18

716 4, 703 2, 703 15, 706920, 706 47

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.

6892366, May 10, 2005

Jan 31, 2002

10/066326

Steven Teig - Menlo Park CA, US
Arindam Chatterjee - San Carlos CA, US

Cadence Design Systems, Inc. - San Jose CA

G06F009/45

716 5, 716 4, 716 7

A system for using machine learning based upon Bayesian inference using a hybrid monte carlo method 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, complex mathematical 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. Next, the system uses Bayesian inference implemented with a Monte Carlo method to train a set of neural networks for extraction problems.

6907591, Jun 14, 2005

Jan 31, 2002

10/062184

Steven Teig - Menlo Park CA, US
Arindam Chatterjee - San Carlos CA, US

Cadence Design Systems, Inc. - San Jose CA

G06F009/45

716 7

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. Next, the system trains a set of neural networks using the training sets. In one embodiment, Bayesian inference is used to train the neural networks that are used to model the extraction.