Einzeltreffer — DigiBib

Approaches for accurate neural network training for library-based critical dimension (CD) metrology are described. Approaches for fast neural network training for library-based CD metrology are also described.

Titel:	Accurate and Fast Neural network Training for Library-Based Critical Dimension (CD) Metrology
Autor/in / Beteiligte Person:	Jin, Wen ; Vuong, Vi ; Bao, Junwei ; Lee, Lie-Quan ; Poslavsky, Leonid
Link:	Zum Volltext
Veröffentlichung:	2012
Medientyp:	Patent
Sonstiges:	Nachgewiesen in: USPTO Patent Applications Sprachen: English Document Number: 20120226644 Publication Date: September 6, 2012 Appl. No: 13/041253 Application Filed: March 04, 2011 Claim: 1. A method of accurate neural network training for library-based critical dimension (CD) metrology, the method comprising: optimizing a threshold for a principal component analysis (PCA) of a spectrum data set to provide a principal component (PC) value; estimating a training target for one or more neural networks; training, based both on the training target and on the PC value provided from optimizing the threshold for the PCA, the one or more neural networks; and providing a spectral library based on the one or more trained neural networks. Claim: 2. The method of claim 1, wherein optimizing the threshold for the PCA comprises determining a lowest level spectrum domain. Claim: 3. The method of claim 1, wherein optimizing the threshold for the PCA of the spectrum data set comprises: determining a first PCA threshold; applying the PCA to the spectrum data set using the first PCA threshold; calculating a spectrum error introduced by applying the PCA using the first PCA threshold; and comparing the spectrum error to a spectrum noise level. Claim: 4. The method of claim 3, further comprising: if the spectrum error is less than the spectrum noise level, setting the first PCA threshold to the PC value. Claim: 5. The method of claim 3, further comprising: if the spectrum error is greater than or equal to than the spectrum noise level, determining a second PCA threshold, and repeating the applying, the calculating, and the comparing. Claim: 6. The method of claim 1, wherein optimizing the threshold for the PCA comprises using a Mueller domain error tolerance. Claim: 7. The method of claim 1, wherein the high accuracy spectral library comprises a simulated spectrum, the method further comprising: comparing the simulated spectrum to a sample spectrum. Claim: 8. A machine-accessible storage medium having instructions stored thereon which cause a data processing system to perform a method of accurate neural network training for library-based critical dimension (CD) metrology, the method comprising: optimizing a threshold for a principal component analysis (PCA) of a spectrum data set to provide a principal component (PC) value; estimating a training target for one or more neural networks; training, based both on the training target and on the PC value provided from optimizing the threshold for the PCA, the one or more neural networks; and providing a spectral library based on the one or more trained neural networks. Claim: 9. The storage medium as in claim 8, wherein optimizing the threshold for the PCA comprises determining a lowest level spectrum domain. Claim: 10. The storage medium as in claim 8, wherein optimizing the threshold for the PCA of the spectrum data set comprises: determining a first PCA threshold; applying the PCA to the spectrum data set using the first PCA threshold; calculating a spectrum error introduced by applying the PCA using the first PCA threshold; and comparing the spectrum error to a spectrum noise level. Claim: 11. The storage medium as in claim 10, the method further comprising: if the spectrum error is less than the spectrum noise level, setting the first PCA threshold to the PC value. Claim: 12. The storage medium as in claim 10, the method further comprising: if the spectrum error is greater than or equal to than the spectrum noise level, determining a second PCA threshold, and repeating the applying, the calculating, and the comparing. Claim: 13. The storage medium as in claim 8, wherein optimizing the threshold for the PCA comprises using a Mueller domain error tolerance. Claim: 14. The storage medium as in claim 8, wherein the high accuracy spectral library comprises a simulated spectrum, the method further comprising: comparing the simulated spectrum to a sample spectrum. Claim: 15. A method of fast neural network training for library-based critical dimension (CD) metrology, the method comprising: providing a training target for a first neural network; training the first neural network, the training comprising starting with a predetermined number of neurons and iteratively increasing the number of neurons until an optimized total number of neurons is reached; generating a second neural network based on the training and the optimized total number of neurons; and providing a spectral library based on the second neural network. Claim: 16. The method of claim 15, wherein iteratively increasing the number of neurons until the optimized total number of neurons is reached comprises using a modified Levenberg-Marquardt approach. Claim: 17. The method of claim 15, wherein iteratively increasing the number of neurons comprises increasing the number of neurons in a hidden layer of the first neural network. Claim: 18. The method of claim 15, wherein the spectral library comprises a simulated spectrum, the method further comprising: comparing the simulated spectrum to a sample spectrum. Claim: 19. A machine-accessible storage medium having instructions stored thereon which cause a data processing system to perform a method of fast neural network training for library-based critical dimension (CD) metrology, the method comprising: providing a training target for a first neural network; training the first neural network, the training comprising starting with a predetermined number of neurons and iteratively increasing the number of neurons until an optimized total number of neurons is reached; generating a second neural network based on the training and the optimized total number of neurons; and providing a spectral library based on the second neural network. Claim: 20. The storage medium as in claim 19, wherein iteratively increasing the number of neurons until the optimized total number of neurons is reached comprises using a modified Levenberg-Marquardt approach. Claim: 21. The storage medium as in claim 19, wherein iteratively increasing the number of neurons comprises increasing the number of neurons in a hidden layer of the first neural network. Claim: 22. The storage medium as in claim 19, wherein the spectral library comprises a simulated spectrum, the method further comprising: comparing the simulated spectrum to a sample spectrum. Current U.S. Class: 706/19 Current International Class: 06

Klicken Sie ein Format an und speichern Sie dann die Daten oder geben Sie eine Empfänger-Adresse ein und lassen Sie sich per Email zusenden.

BibTeX Citavi, JabRef, u.a.
(Literaturverwaltung)

PDF kein Volltext!
(Merkzettel, Notizen)

RIS Endnote, Citavi u.a.
(Literaturverwaltung)

MODS
(XML zur Weiterverarbeitung)

oder

Wählen Sie das für Sie passende Zitationsformat und kopieren Sie es dann in die Zwischenablage, lassen es sich per Mail zusenden oder speichern es als PDF-Datei.

Gewünschter Zitations-Stil:

oder

Bitte prüfen Sie, ob die Zitation formal korrekt ist, bevor Sie sie in einer Arbeit verwenden. Benutzen Sie gegebenenfalls den "Exportieren"-Dialog, wenn Sie ein Literaturverwaltungsprogramm verwenden und die Zitat-Angaben selbst formatieren wollen.

Accurate and Fast Neural network Training for Library-Based Critical Dimension (CD) Metrology