Machine learning-based down selection of candidate hotspot locations of circuit designs
A method may include the steps of accessing an input data set of hotspot locations on manufactured circuits of a circuit design. The hotspot locations may be confirmed through a high precision imaging process from a set of candidate locations of the circuit design determined by a low precision imaging process. The method may further include correlating the hotspot locations to layout data for the circuit design, extracting fragment feature vectors for the hotspot locations from optical proximity correction (OPC) fragments of the layout data, processing the fragment feature vectors, providing the processed fragment feature vectors as a training set for training a machine-learning model, and applying the machine-learning model to down select a different set of candidate locations determined by the low precision imaging process.
1 . A method comprising:
by a computing system:
accessing an input data set of hotspot locations on manufactured circuits of a circuit design, wherein the hotspot locations were confirmed through a high precision imaging process from a set of candidate locations of the circuit design determined by a low precision imaging process;
correlating the hotspot locations to layout data for the circuit design;
extracting fragment feature vectors for the hotspot locations from optical proximity correction (OPC) fragments of the layout data, including hotspot fragment feature vectors and non-hotspot fragment feature vectors for the hotspot locations;
processing the fragment feature vectors such that the hotspot fragment feature vectors are a threshold percentage of the total number of feature vectors in the fragment feature vectors;
providing the processed fragment feature vectors as a training set for training a machine-learning model; and
applying the machine-learning model to down select a different set of candidate locations determined by the low precision imaging process, including by
analyzing layout data correlated to the candidate locations based on an uncertainty range of the low precision imaging process to determine candidate fragments;
using the machine-learning model to determine a defect likelihood for the candidate fragments; and
down sampling the different set of candidate locations based on the determined defect likelihoods of the candidate fragments.
2 . The method of claim 1 , wherein the input data set further comprises non-hotspot locations of the circuit design, and further comprising:
correlating the non-hotspot locations to the layout data; and
extracting non-hotspot fragment feature vectors from the layout data that correspond to the non-hotspot locations.
3 . The method of claim 1 , wherein extracting comprises, for a given hotspot location, determining an extraction window for the layout data based on a position of the given hotspot location in the circuit design and an uncertainty range of an imaging technique used to detect the given hotspot location in a manufactured circuit.
4 . The method of claim 1 , wherein extracting further comprises, for a given hotspot location:
determining, as hotspot fragments, OPC fragments within a threshold distance range from the given hotspot location; and
extracting hotspot fragment feature vectors for the given hotspot location from the hotspot fragments.
5 . The method of claim 4 , wherein extracting further comprises, for the given hotspot location:
identifying uncertainty fragments on a same polygon as at least one of the determined hotspot fragments;
determining, as non-hotspot fragments, OPC fragments not within the threshold distance range from the given hotspot location and not identified as the uncertainty fragments; and
extracting non-hotspot fragment feature vectors for the given hotspot location from the non-hotspot fragments.
6 . The method of claim 1 , wherein processing the fragment feature vectors comprises data boosting the hotspot fragment feature vectors to the threshold percentage of the total number of feature vectors in the fragment feature vectors.
7 . The method of claim 6 , wherein processing the fragment feature vectors further comprises grouping the hotspot fragment feature vectors according to hotspot characteristics prior to data boosting the hotspot fragment feature vectors, and
wherein the data boosting of the hotspot fragment feature vectors is performed according to balancing criteria applied for the grouped hotspot fragment feature vectors.
8 . The method of claim 1 , wherein the fragment feature vectors comprise a vector portion that characterizes manufacturing process characteristics of the hotspot locations.
9 . The method of claim 1 , wherein the fragment feature vectors comprise a vector portion that characterizes particular candidate locations from which the hotspot locations were confirmed.
10 . A non-transitory computer-readable medium comprising instructions that, when executed by a processor, cause a computing system to:
access an input data set of hotspot locations on manufactured circuits of a circuit design, wherein the hotspot locations were confirmed through a high precision imaging process from a set of candidate locations of the circuit design determined by a low precision imaging process;
correlate the hotspot locations to layout data for the circuit design;
extract fragment feature vectors for the hotspot locations from optical proximity correction (OPC) fragments of the layout data, including hotspot fragment feature vectors and non-hotspot fragment feature vectors for the hotspot locations;
process the fragment feature vectors such that the hotspot fragment feature vectors are a threshold percentage of the total number of feature vectors in the fragment feature vectors;
provide the processed fragment feature vectors as a training set for training a machine-learning model; and
apply the machine-learning model to down select a different set of candidate locations determined by the low precision imaging process,
wherein the instructions cause the computing system to process the fragment feature vectors by data boosting the hotspot fragment feature vectors to the threshold percentage of the total number of feature vectors in the fragment feature vectors,
wherein instructions cause the computing system to process the fragment feature vectors further by grouping the hotspot fragment feature vectors according to hotspot characteristics prior to the data boosting of the hotspot fragment feature vectors, and
wherein the data boosting of the hotspot fragment feature vectors is performed according to balancing criteria applied for the grouped hotspot fragment feature vectors.
11 . The non-transitory computer-readable medium of claim 10 , wherein the input data set-further comprises non-hotspot locations of the circuit design, and wherein the instructions further cause the computing system to:
correlate the non-hotspot locations to the layout data; and
extract non-hotspot fragment feature vectors from the layout data that correspond to the non-hotspot locations.
12 . The non-transitory computer-readable medium of claim 10 , wherein the instructions cause the computing system to extract the fragment feature vectors by, for a given hotspot location, determining an extraction window for the layout data based on a position of the given hotspot location in the circuit design and an uncertainty range of an imaging technique used to detect the given hotspot location in a manufactured circuit.
13 . The non-transitory computer-readable medium of claim 10 , wherein the instructions cause the computing system to extract the fragment feature vectors further by, for a given hotspot location:
determining, as hotspot fragments, OPC fragments within a threshold distance range from the given hotspot location; and
extracting hotspot fragment feature vectors for the given hotspot location from the hotspot fragments.
14 . The non-transitory computer-readable medium of claim 13 , wherein the instructions cause the computing system to extract the fragment feature vectors further by, for the given hotspot location:
identifying uncertainty fragments on a same polygon as at least one of the determined hotspot fragments;
determining, as non-hotspot fragments, OPC fragments not within the threshold distance range from the given hotspot location and not identified as the uncertainty fragments; and
extracting non-hotspot fragment feature vectors for the given hotspot location from the non-hotspot fragments.
15 . The non-transitory computer-readable medium of claim 10 , wherein the instructions cause the computing system to apply the machine-learning model by:
analyzing layout data correlated to the candidate locations based on an uncertainty range of the low precision imaging process to determine candidate fragments;
using the machine-learning model to determine a defect likelihood for the candidate fragments; and
down sampling the different set of candidate locations based on the determined defect likelihoods of the candidate fragments.
16 . The non-transitory computer-readable medium of claim 10 , wherein the fragment feature vectors comprise a vector portion that characterizes manufacturing process characteristics of the hotspot locations.
17 . The non-transitory computer-readable medium of claim 10 , wherein the fragment feature vectors comprise a vector portion that characterizes particular candidate locations from which the hotspot locations were confirmed.
18 . A non-transitory computer-readable medium comprising instructions that, when executed by a processor, cause a computing system to:
access an input data set of hotspot locations on manufactured circuits of a circuit design, wherein the hotspot locations were confirmed through a high precision imaging process from a set of candidate locations of the circuit design determined by a low precision imaging process;
correlate the hotspot locations to layout data for the circuit design;
extract fragment feature vectors for the hotspot locations from optical proximity correction (OPC) fragments of the layout data, including hotspot fragment feature vectors and non-hotspot fragment feature vectors for the hotspot locations;
process the fragment feature vectors such that the hotspot fragment feature vectors are a threshold percentage of the total number of feature vectors in the fragment feature vectors;
provide the processed fragment feature vectors as a training set for training a machine-learning model; and
apply the machine-learning model to down select a different set of candidate locations determined by the low precision imaging process, including by
analyzing layout data correlated to the candidate locations based on an uncertainty range of the low precision imaging process to determine candidate fragments;
using the machine-learning model to determine a defect likelihood for the candidate fragments; and
down sampling the different set of candidate locations based on the determined defect likelihoods of the candidate fragments.
19 . The non-transitory computer-readable medium of claim 18 , wherein the input data set further comprises non-hotspot locations of the circuit design, and wherein the instructions further cause the computing system to:
correlate the non-hotspot locations to the layout data; and
extract non-hotspot fragment feature vectors from the layout data that correspond to the non-hotspot locations.
20 . The non-transitory computer-readable medium of claim 18 , wherein the instructions cause the computing system to extract the fragment feature vectors further by, for a given hotspot location:
determining, as hotspot fragments, OPC fragments within a threshold distance range from the given hotspot location;
extracting hotspot fragment feature vectors for the given hotspot location from the hotspot fragments;
identifying uncertainty fragments on a same polygon as at least one of the determined hotspot fragments;
determining, as non-hotspot fragments, OPC fragments not within the threshold distance range from the given hotspot location and not identified as the uncertainty fragments; and
extracting non-hotspot fragment feature vectors for the given hotspot location from the non-hotspot fragments.