Training spectrum generation for machine learning system for spectrographic monitoring
A method of generating training spectra for training of a neural network includes measuring a first plurality of training spectra from one or more sample substrates, measuring a characterizing value for each training spectra of the plurality of training spectra to generate a plurality of characterizing values with each training spectrum having an associated characterizing value, measuring a plurality of dummy spectra during processing of one or more dummy substrates, and generating a second plurality of training spectra by combining the first plurality of training spectra and the plurality of dummy spectra, there being a greater number of spectra in the second plurality of training spectra than in the first plurality of training spectra. Each training spectrum of the second plurality of training spectra having an associated characterizing value.
1. A method of generating training spectra for training of a neural network, comprising:
measuring a first plurality of training spectra from one or more sample substrates;
measuring a characterizing value for each training spectra of the first plurality of training spectra to generate a plurality of characterizing values with each training spectrum having an associated characterizing value; and
generating a second plurality of training spectra by adding a plurality of different noise components to the first plurality of training spectra, there being a greater number of spectra in the second plurality of training spectra than in the first plurality of training spectra, each training spectrum of the second plurality of training spectra having an associated characterizing value.
2. The method of claim 1 , wherein generating the second plurality of training spectra includes adding a plurality of different noise components to each training spectrum from the first plurality of training spectra.
3. The method of claim 1 , comprising measuring a plurality of dummy spectra during processing of one or more dummy substrates.
4. The method of claim 3 , wherein adding the plurality of different noise components includes by combining the first plurality of training spectra and the plurality of dummy spectra.
5. The method of claim 3 , wherein the dummy substrates comprises blank semiconductor substrates.
6. The method of claim 3 , wherein the processing comprises chemical mechanical polishing.
7. A computer program product for generating spectra for training of a neural network, the computer program product tangibly embodied in a non-transitory computer readable media and comprising instructions for causing a processor to:
receive a first plurality of training spectra and plurality of characterizing values, each training spectrum of the first plurality of training spectra representing a spectrum from a substrate and having an associated characterizing value for the substrate from the plurality of characterizing values; and
generate a second plurality of training spectra by adding a plurality of different noise components to the first plurality of training spectra, there being a greater number of spectra in the second plurality of training spectra than in the first plurality of training spectra, each training spectrum of the second plurality of training spectra having an associated characterizing value.
8. The computer program product of claim 7 , wherein the instructions to generate the second plurality of training spectra include instructions to add a plurality of different noise components to each training spectrum from the first plurality of training spectra.
9. The computer program product of claim 7 , comprising receive a plurality of dummy spectra representing spectra obtained from processing of one or more dummy substrates.
10. The computer program product of claim 9 , wherein the instructions to add the plurality of different noise components include instructions to combine the first plurality of training spectra and the plurality of dummy spectra.
11. The computer program product of claim 9 , comprising instructions to normalize the plurality of dummy spectra to generate a plurality of normalized dummy spectra.
12. The computer program product of claim 11 , wherein the instructions for combining the first plurality of training spectra and the plurality of dummy spectra comprise instructions to, for each normalized dummy spectrum from the plurality of normalized dummy spectra, multiply the normalized dummy spectrum by one of the first plurality of training spectra to generate one of the second plurality of training spectra.
13. The computer program product of claim 12 , comprising instructions to multiply the normalized dummy spectrum by a randomly selected one of the first plurality of training spectra.
14. The computer program product of claim 7 , wherein the characterizing value comprises a thickness of an outermost layer of the substrate.