IEEE transactions on systems, man, and cybernetics, Jun 1, 1998
In this paper, fuzzy inference models for pattern classifications have been developed and fuzzy i... more In this paper, fuzzy inference models for pattern classifications have been developed and fuzzy inference networks based on these models are proposed. Most of the existing fuzzy rule-based systems have difficulties in deriving inference rules and membership functions directly from training data. Rules and membership functions are obtained from experts. Some approaches use backpropagation (BP) type learning algorithms to learn the parameters of membership functions from training data. However, BP algorithms take a long time to converge and they require an advanced setting of the number of inference rules. The work to determine the number of inference rules demands lots of experiences from the designer. In this paper, self-organizing learning algorithms are proposed for the fuzzy inference networks. In the proposed learning algorithms, the number of inference rules and the membership functions in the inference rules will be automatically determined during the training procedure. The learning speed is fast. The proposed fuzzy inference network (FIN) classifiers possess both the structure and the learning ability of neural networks, and the fuzzy classification ability of fuzzy algorithms. Simulation results on fuzzy classification of two-dimensional data are presented and compared with those of the fuzzy ARTMAP. The proposed fuzzy inference networks perform better than the fuzzy ARTMAP and need less training samples.
A novel processing platform for post tape out flows
As the computational requirements for post tape out (PTO) flows increase at the 7nm and below tec... more As the computational requirements for post tape out (PTO) flows increase at the 7nm and below technology nodes, there is a need to increase the scalability of the computational tools in order to reduce the turn-around time (TAT) of the flows. Utilization of design hierarchy has been one proven method to provide sufficient partitioning to enable PTO processing. However, as the data is processed through the PTO flow, its effective hierarchy is reduced. The reduction is necessary to achieve the desired accuracy. Also, the sequential nature of the PTO flow is inherently non-scalable. To address these limitations, we are proposing a quasi-hierarchical solution that combines multiple levels of parallelism to increase the scalability of the entire PTO flow. In this paper, we describe the system and present experimental results demonstrating the runtime reduction through scalable processing with thousands of computational cores.
1995 International Conference on Acoustics, Speech, and Signal Processing
A fuzzy inference network (FIN) is proposed. The proposed FIN preserves the advantages of both fu... more A fuzzy inference network (FIN) is proposed. The proposed FIN preserves the advantages of both fuzzy classification algorithm and neural networks. It can learn membership functions directly from training samples and classify patterns according to the membership values. As efficient self-organizing learning algorithm is also presented.
Supervised fuzzy inference network for invariant pattern recognition
Midwest Symposium on Circuits and Systems, 2000
A supervised fuzzy inference network (FIN) model and its learning algorithm for invariant pattern... more A supervised fuzzy inference network (FIN) model and its learning algorithm for invariant pattern recognition are presented in this paper. This fuzzy inference network is suitable for 2-D visual pattern recognition problems and has been tested with letter patterns of black and white pixel values. In contrast to most of the conventional pattern recognition systems, the proposed fuzzy inference network
Simultaneous layout, process, and model optimization within an integrated design-for-yield environment
SPIE Proceedings, 2006
Trends in the design feature shrinking that outrun the progress in the lithography technologies r... more Trends in the design feature shrinking that outrun the progress in the lithography technologies require critical efforts in the layout, process, and model development. Printing a layout is no longer a problem only for the lithographers; it has penetrated into the layout stage as well. Layout patterns are getting more aggressive, raising serious printability concerns. This requires very accurate models
A fuzzy neural classifier for pattern classificatoin
1993 IEEE International Symposium on Circuits and Systems
A novel fuzzy neural classifier (FNC) and its learning algorithm are proposed. This FNC can learn... more A novel fuzzy neural classifier (FNC) and its learning algorithm are proposed. This FNC can learn the membership function of each fuzzy class from training samples. The learning speed of the FNC is fast. Simulation results on the learning and estimation of the membership functions of one-dimensional samples are presented
IEEE Transactions on Systems, Man and Cybernetics, Part B (Cybernetics), 1998
In this paper, fuzzy inference models for pattern classifications have been developed and fuzzy i... more In this paper, fuzzy inference models for pattern classifications have been developed and fuzzy inference networks based on these models are proposed. Most of the existing fuzzy rule-based systems have difficulties in deriving inference rules and membership functions directly from training data. Rules and membership functions are obtained from experts. Some approaches use backpropagation (BP) type learning algorithms to learn the parameters of membership functions from training data. However, BP algorithms take a long time to converge and they require an advanced setting of the number of inference rules. The work to determine the number of inference rules demands lots of experiences from the designer. In this paper, self-organizing learning algorithms are proposed for the fuzzy inference networks. In the proposed learning algorithms, the number of inference rules and the membership functions in the inference rules will be automatically determined during the training procedure. The learning speed is fast. The proposed fuzzy inference network (FIN) classifiers possess both the structure and the learning ability of neural networks, and the fuzzy classification ability of fuzzy algorithms. Simulation results on fuzzy classification of two-dimensional data are presented and compared with those of the fuzzy ARTMAP. The proposed fuzzy inference networks perform better than the fuzzy ARTMAP and need less training samples.
In this paper, we define four types of fuzzy neurons and propose the structure of a four-layer fe... more In this paper, we define four types of fuzzy neurons and propose the structure of a four-layer feedforward fuzzy neural network (FNN) and its associated learning algorithm. The proposed four-layer FNN performs well when used to recognize shifted and distorted training patterns. When an input pattern is provided, the network first fuzzifies this pattern and then computes the similarities of this pattern to all of the learned patterns. The network then reaches a conclusion by selecting the learned pattern with the highest similarity and gives a nonfuzzy output. The 26 English alphabets and the 10 Arabic numerals, each represented by 16x 16 pixels, were used as original training patterns. In the simulation experiments, the original 36 exemplar patterns were shifted in eight directions by 1 pixel (6.25% to 8.84%) and 2 pixels (12.5'h to 17.68% ). After the FNN has been trained by the 36 exemplar patterns, the FNN can recall all of the learned patterns with recognition rate. It can also recognize patterns shifted by 1 pixel in eight directions with loo%, recognition rate and patterns shifted by 2 pixels in eight directions with an average recognition rate of 92.01%. After the FNN has been trained by the 36 exemplar patterns and 72 shifted patterns, it can recognize patterns shifted by 1 pixel with recognition rate and patterns shifted by 2 pixels with an average recognition rate of 98.61%. We have also tested the FNN with 10 kinds of distorted patterns for each of the 36 exemplars. The FNN can recognize all of the distorted patterns with 100% recognition rate. The proposed FNN can also be adapted for applications in some other pattern recognition problems. A structure which is composed of many processing elements connected to each other through weights [ 11-[3]. Neural networks (NN's) are built after biological neural systems. A NN stores patterns with distributed coding and is a trainable nonlinear dynamic system. A NN has a faster response and a higher performance than those of a sequential digital computer in emulating the capabilities of the human brain. Recently, NN's have been used in pattem recognition problems, especially where input patterns are shifted in position and scale-changed. Fukushima et al. [4], [SI have presented the Neocognitron, which is insensitive to translation and deformation of input patterns, and used it to recognize handprinted characters. However, the Neocognitron is complex Manuscript
Dimensions for 32nm generation logic are expected to be ~45nm. Even with high NA scanners, the k ... more Dimensions for 32nm generation logic are expected to be ~45nm. Even with high NA scanners, the k 1 factor is below 0.32. Gridded-design-rules (GDR) are a form of restricted design rules (RDR) and have a number of benefits from design through fabrication. The combination of rules and topologies can be verified during logic technology development, much as is done with memories. Topologies which have been preverified can be used to implement random logic functions with "hotspot" prevention that is virtually context-independent. Mask data preparation is simplified with less aggressive OPC, resulting in shorter fracturing, writing, and inspection times. In the wafer fab, photolithography, etch, and CMP are more controllable because of the grating-like patterns. Tela Canvas™ GDR layout was found to give smaller area cells than a conventional 2D layout style. Variability and context independence were also improved.
Defect printability analysis study using virtual stepper system in a production environment
Proceedings of SPIE, Jul 1, 2002
In this paper the simulation of wafer images for Attenuated Phase Shift Masks (ATTPSM) and repair... more In this paper the simulation of wafer images for Attenuated Phase Shift Masks (ATTPSM) and repaired binary masks are performed by Virtual Stepper System in a real production environment. In addition, the Automatic Defect Severity Scoring module in Virtual Stepper is also used to calculate the defect severity score for each defect. ADSS provides an overall score that quantifies the
Dimensions for 32nm generation logic are expected to be ~45nm. Even with high NA scanners, the k ... more Dimensions for 32nm generation logic are expected to be ~45nm. Even with high NA scanners, the k 1 factor is below 0.32. Gridded-design-rules (GDR) are a form of restricted design rules (RDR) and have a number of benefits from design through fabrication. The combination of rules and topologies can be verified during logic technology development, much as is done with memories. Topologies which have been preverified can be used to implement random logic functions with "hotspot" prevention that is virtually context-independent. Mask data preparation is simplified with less aggressive OPC, resulting in shorter fracturing, writing, and inspection times. In the wafer fab, photolithography, etch, and CMP are more controllable because of the grating-like patterns. Tela Canvas™ GDR layout was found to give smaller area cells than a conventional 2D layout style. Variability and context independence were also improved.
<title>Automated Defect Severity Analysis for Binary and PSM Mask Defects</title>
22nd Annual BACUS Symposium on Photomask Technology, 2002
Traditionally, mask defect analysis has been done through a visual inspection review. As the semi... more Traditionally, mask defect analysis has been done through a visual inspection review. As the semiconductor industry moves into smaller process generations and the complexity of mask exponentially increases, the traditional mask defect analysis method becomes very time consuming. The Automatic Defect Severity Scoring (ADSS) module of i-Virtual Stepper System from Numerical Technologies offers an extremely fast and highly accurate software solution for defect printability analysis of advanced masks such as OPC and phase-shifting masks in a real production environment. In a previous paper [1], we have introduced the ADSS concept and discussed some results for line-space patterns on OPC and non-OPC masks. In this paper, we will discuss the ADSS results for both line-space and contact patterns on attenuated phase-shifting masks (ATTPSM), together with some ADSS results for line-space patterns on binary masks. The ADSS results are compared to wafer results. The wafer exposures were performed using 248 nm imaging technology and inspection images were generated on a KLA-Tencor’s SLF27 system.
Simulation-based mask quality control in a production environment
Metrology, Inspection, and Process Control for Microlithography XVIII, 2004
Traditionally, mask defect analysis has been done through a visual inspection review. As the semi... more Traditionally, mask defect analysis has been done through a visual inspection review. As the semiconductor industry moves into smaller process generations and the complexity of mask exponentially increases, &amp;amp;amp;amp;amp;amp;quot;Mask&amp;amp;amp;amp;amp;amp;quot; issues have emerged as one of the main production problems due to their rising cost and long turn-around time. Mask-making specifications related to defects found on advanced masks also becomes more
Enhanced dispositioning of reticle defects using the Virtual Stepper with automated defect severity scoring
Photomask and Next-Generation Lithography Mask Technology VIII, 2001
Enhanced dispositioning of reticle defects using the Virtual Stepper with automated defect severi... more Enhanced dispositioning of reticle defects using the Virtual Stepper with automated defect severity scoring. [Proceedings of SPIE 4409, 467 (2001)]. Lynn Cai, Khoi A. Phan, Chris A. Spence, Linyong Pang, Kevin K. Chan. Abstract. ...
<title>Automatic defect severity scoring for 193-nm reticle defect inspection</title>
Optical Microlithography XIV, 2001
Sub-wavelength lithography requires knowledgeable application of resolution enhancement technique... more Sub-wavelength lithography requires knowledgeable application of resolution enhancement techniques (RETs) such as optical proximity correction (OPC) and phase shift mask (PSM). Use of RETs, in turn, requires that new photomask specifications and special requirements for mask defect printability be taken into consideration. This is especially true, as the photomask's critical dimensions become more aggressive (400 nm moving toward 300 nm). Traditionally, mask defect analysis and subsequent defect disposition has been accomplished by first performing automated reticle inspection, and then by visual inspection ultimately dependent on operator judgement. As the semiconductor industry moves to more challenging process generations this methodology is no longer viable for assessing the impact of a defect on the printed wafer. New techniques for more accurate, production-worthy defect printability analysis and defect disposition procedures are required. Developed at Numerical Technologies, Inc. is the Virtual StepperTM System that offers a fast, accurate software solution for defect printability analysis based on state-of- the-art lithography simulation techniques for advanced masks production using OPC and PSM. The newly developed Virtual Stepper System feature, Automatic Defect Severity Scoring (ADSS) provides fully automated and accurate defect impact analysis capability by calculating a consistent Defect Severity Score (DSS) for each defect detected by an inspection tool. DSS is an overall score that quantifies the impact of a given defect on surrounding features and can be used as a comprehensive indicator of defect printability. Taken into consideration, are not only printing defects, but defects which cause critical dimension (CD) errors altering a given process window.
<title>Enhanced dispositioning of reticle defects for advanced masks using virtual stepper with automated defect severity scoring</title>
23rd Annual BACUS Symposium on Photomask Technology, 2003
As the semiconductor industry continues to scale down critical dimensions (CD), proximity effects... more As the semiconductor industry continues to scale down critical dimensions (CD), proximity effects get more and more severe. As such, aggressive Optical Proximity Correction (OPC) features like hammerheads, serifs and assist bars inevitably appear on fabricated masks. The great challenge, however -- to reliably assure the quality of these advanced masks -- is to be able to directly judge a
A novel processing platform for post tape out flows
Optical Microlithography XXXI, 2018
As the computational requirements for post tape out (PTO) flows increase at the 7nm and below tec... more As the computational requirements for post tape out (PTO) flows increase at the 7nm and below technology nodes, there is a need to increase the scalability of the computational tools in order to reduce the turn-around time (TAT) of the flows. Utilization of design hierarchy has been one proven method to provide sufficient partitioning to enable PTO processing. However, as the data is processed through the PTO flow, its effective hierarchy is reduced. The reduction is necessary to achieve the desired accuracy. Also, the sequential nature of the PTO flow is inherently non-scalable. To address these limitations, we are proposing a quasi-hierarchical solution that combines multiple levels of parallelism to increase the scalability of the entire PTO flow. In this paper, we describe the system and present experimental results demonstrating the runtime reduction through scalable processing with thousands of computational cores.
Uploads
Papers by Lynn Cai