Data-mining-based knowledge extraction and visualization of analog/mixed-signal/custom digital circuit design flow
View Patent ↗A system and method of generating a set of circuit simulation data, applying data mining to for knowledge extraction from the data, and graphically presenting the extracted knowledge in a format that is easy to digest to a designer.
1. A tangible computer-readable medium having recorded thereon statements and instructions for execution by a computer of a method for visualizing data associated with an electrical circuit design (ECD), the ECD having associated therewith circuit design variables and circuit performance metrics, the method comprising steps of:
accessing a database containing circuit simulation data of the ECD, the circuit simulation data including circuit design variables data and circuit performance metrics data;
extracting circuit simulation data to obtain extracted data;
processing the extracted data in accordance with a pre-determined data-mining algorithm to obtain output data, the output data correlating at least one of:
(a) a design variable to a circuit performance metric; and
(b) a first performance metric to a second performance metric;
organizing the output data in accordance with pre-determined display criteria to obtain organized data; and
displaying the organized data;
wherein processing the extracted data includes forming a classification and regression tree (CART), the CART mapping at least one circuit variable to one of the circuit performance metrics.
2. The tangible computer-readable of claim 1 wherein the step of accessing the database is preceded with a step of populating the database.
3. The tangible computer-readable of claim 2 wherein the step of populating the database is performed using SPICE simulation of the ECD.
4. The tangible computer-readable of claim 2 wherein the step of populating the database includes simulating the ECD by using Latin hypercube samples of a design variable space of the ECD.
5. A tangible computer-readable medium having recorded thereon statements and instructions for execution by a computer of a method for visualizing data associated with an electrical circuit design (ECD), the ECD having associated therewith circuit design variables and circuit performance metrics, the method comprising steps of:
accessing a database containing circuit simulation data of the ECD, the circuit simulation data including circuit design variables data and circuit performance metrics data;
extracting circuit simulation data to obtain extracted data;
processing the extracted data in accordance with a pre-determined data-mining algorithm to obtain output data, the output data correlating at least one of:
(a) a design variable to a circuit performance metric; and
(b) a first performance metric to a second performance metric;
organizing the output data in accordance with pre-determined display criteria to obtain organized data; and
displaying the organized data;
wherein the step of processing the extracted data includes a step of building a regression model representing a circuit performance metric as a function of at least one circuit variable.
6. The tangible computer-readable of claim 5 wherein the step of building a regression model is followed by steps of:
selecting input data from the circuit variables data to obtain input data;
inputting the input data into the regression model to obtain an output of the regression model; and
comparing the output of the regression model to circuit performance metrics data associated with the input data, to obtain a comparison value.
7. The tangible computer-readable of claim 6 wherein if the comparison value is outside a pre-determined range, the steps of selecting input data, inputting the input data and comparing the output are followed by a step of modifying the regression model in accordance with the comparison value, and are repeated until the comparison value is within the pre-determined range.
8. The tangible computer-readable of claim 5 wherein processing the extracted data includes performing a step of clustering the extracted data according to a pre-determined distance metric criteria.
9. The tangible computer-readable of claim 5 wherein processing the extracted data includes forming at least one of a set of classification and regression trees (CARTs) and a Bayes network.
10. A system for visualizing data associated with an electrical circuit design (ECD), the ECD having circuit variables and circuit performance metrics, the system comprising:
a database containing circuit simulation data of the ECD, the circuit simulation data including circuit variables data and circuit performance metrics data;
a data-mining engine for extracting circuit simulation data to obtain extracted data and to process the extracted data in accordance with a pre-determined data-mining algorithm to obtain output data, said data mining engine configured to process the extracted data by building a regression model representing a circuit performance metric as a function of at least one circuit variable such that the output data correlating at least one of (a) a design variable to a circuit performance metric, and (b) a first performance metric to a second performance metric;
a visualization tool for organizing the output data in accordance with pre-determined display criteria to obtain organized data; and
a display system for displaying the organized data.
11. The system as claimed in claim 10 wherein said data mining engine if further configured to process the extracted data by forming a classification and regression tree (CART), the CART mapping at least one circuit variable to one of the circuit performance metrics.