Methods and apparatuses for calibrating a sensor
A method for determining a calibration function includes: calculating a first distance, between a distribution of target spectra and a comparison distribution of spectra; calibrating the distribution of target spectra with a first preliminary calibration function to form a first distribution of calibrated target spectra; calculating a second distance, between the first distribution of calibrated target spectra and the comparison distribution of spectra; determining that the second distance is less than the first distance; and setting the calibration function equal to the first preliminary calibration function.
1. A method for determining a calibration function, the method comprising:
calculating a first distance, between a distribution of target spectra and a comparison distribution of spectra;
calibrating the distribution of target spectra with a first preliminary calibration function to form a first distribution of calibrated target spectra;
calculating a second distance, between the first distribution of calibrated target spectra and the comparison distribution of spectra;
determining, after calculating the second distance, that the second distance is less than the first distance; and setting, after determining that the second distance is less than the first distance, the calibration function equal to the first preliminary calibration function,
wherein the distribution of target spectra comprises one or more spectra measured by a target sensor, and the comparison distribution of spectra comprises a plurality of comparison distributions corresponding to a plurality of respective comparison sensors, each of the comparison distributions comprising one or more spectra measured by the corresponding comparison sensor, and
wherein the target sensor comprises a light source.
2. The method of claim 1 , wherein the target sensor is an optical spectrophotometer comprising the light source, and each of the comparison sensors is an optical spectrophotometer.
3. A method for determining a calibration function, the method comprising:
calculating a first distance, between a distribution of target spectra and a comparison distribution of spectra;
calibrating the distribution of target spectra with a first preliminary calibration function to form a first distribution of calibrated target spectra;
calculating a second distance, between the first distribution of calibrated target spectra and the comparison distribution of spectra;
determining, after calculating the second distance, that the second distance is less than the first distance; and setting, after determining that the second distance is less than the first distance, the calibration function equal to the first preliminary calibration function,
wherein the distribution of target spectra comprises one or more spectra measured by a target sensor, and the comparison distribution of spectra comprises a plurality of comparison distributions corresponding to a plurality of respective comparison sensors, each of the comparison distributions comprising one or more spectra measured by the corresponding comparison sensor, and
wherein the comparison distribution of spectra is based on the plurality of comparison distributions.
4. The method of claim 3 , wherein the comparison distribution of spectra is the union of the plurality of comparison distributions.
5. The method of claim 1 , wherein the target spectra comprise spectra of one or more target samples, each of the one or more target samples being from a same class of samples, and
wherein the spectra of the comparison distribution of spectra are spectra of one or more comparison samples, each of the one or more comparison samples being from the same class of samples as the one or more target samples are from.
6. The method of claim 5 , further comprising:
illuminating each of the one or more target samples with light of a respective plurality of actual wavelengths; and
measuring, by the target sensor and for each of the one or more target samples, respective portions of the light of the respective plurality of actual wavelengths transmitted through the target sample, wherein the target spectra comprise, for each of the one or more target samples, a plurality of absorbance values of the target sample that are respectively based on the measured portions of light.
7. The method of claim 1 , wherein the first distance is calculated based on a metric, and the metric is a Wasserstein distance metric, a Kullback-Leibler divergence metric, a Renyi divergence metric, or an f-divergence metric.
8. The method of claim 1 , further comprising selecting the first preliminary calibration function from among a family of preliminary calibration functions.
9. The method of claim 1 , wherein the first preliminary calibration function maps a set of nominal wavelengths to a set of calibrated wavelengths.
10. The method of claim 9 , further comprising defining the first preliminary calibration function by training a neural network to provide a calibrated wavelength of the set of calibrated wavelengths in response to being provided with a nominal wavelength of the set of nominal wavelengths.
11. A processing circuit for determining a calibration function, the processing circuit being configured to:
calculate a first distance, between a distribution of target spectra and a comparison distribution of spectra;
calibrate the distribution of target spectra with a first preliminary calibration function to form a first distribution of calibrated target spectra;
calculate a second distance, between the first distribution of calibrated target spectra and the comparison distribution of spectra;
determine, after calculating the second distance, that the second distance is less than the first distance; and set, after determining that the second distance is less than the first distance, the calibration function equal to the first preliminary calibration function,
wherein the distribution of target spectra comprises one or more spectra measured by a target sensor, and the comparison distribution of spectra comprises a plurality of comparison distributions corresponding to a plurality of respective comparison sensors, each of the comparison distributions comprising one or more spectra measured by the corresponding comparison sensor, and
wherein the target sensor comprises a light source.
12. The processing circuit of claim 11 , wherein the target sensor is an optical spectrophotometer comprising the light source, and each of the comparison sensors is an optical spectrophotometer.
13. The processing circuit of claim 11 , wherein the comparison distribution of spectra is based on the plurality of comparison distributions.
14. The processing circuit of claim 11 , wherein the processing circuit is in the target sensor.
15. The processing circuit of claim 14 , wherein the processing circuit is further configured:
to receive the distribution of target spectra; and
to receive, from a cloud storage or a memory external to the target sensor, the one or more comparison distributions.
16. The processing circuit of claim 11 , wherein the target spectra comprise spectra of one or more target samples, each of the one or more target samples being from a same class of samples, and
wherein the spectra of the comparison distribution of spectra are spectra of one or more comparison samples, each of the one or more comparison samples being from the same class of samples as the one or more target sample are from.
17. The processing circuit of claim 11 , wherein the first distance is calculated based on a metric, and the metric is a Wasserstein distance metric, a Kullback-Leibler divergence metric, a Renyi divergence metric, or an f-divergence metric.
18. The processing circuit of claim 11 , wherein the first preliminary calibration function maps a set of nominal wavelengths to a set of calibrated wavelengths.