We had a modest amount of high-dimensional data consisting of signals acquired by an inexpensive hand held radiation detector while driving around Chicago. The signal to noise ratio in each signal was very bad. The challenge was to average a few relevant signals to bring out the signature of the radioactive materials of interest, since averaging all signals will reveal nothing of interest. Doing this using conventional techniques presented huge computational challenges.
The results: DQC analysis revealed that, among the large number of clusters present two were interesting. One cluster contained 22 readings and the other, 44 readings. Averaging each of these clusters revealed the presence of Co 60 and Cs 137, the radioactive materials of interest. Averaging the remaining clusters showed nothing of interest. The combinatoric problem of selecting such a small number of readings to average – out of the 7000 samples – would have been prohibitive using conventional approaches.
Conclusion: DQC demonstrated that it can enable the development of relatively inexpensive radiation detectors that can be used by TSA, Homeland Security, police departments, etc. to search for dangerous materials.