Hi James,
Thanks a lot for your question and interest in this work. In short, this is actually resulting from some parameters of the clustering step and the fact that different emission spots may emit light at different time and for longer/shorter duration.
In Fig 6, let's name the spot at (20,10) as spot A and the one at (30, 20) spot B. Spot A emits a moderate quantity of photons over a long period of time (e.g. 1 s, just for illustration purpose) while spot B emits a lot but in a very short time (e.g. 100 ns). When photons are collected over a very long period of time with many iterations of the test loop, Spot A appears brighter thanks to the longer duration of emission. When we applied clustering, parameters were such that it identified mainly the groups of photons that were emitted from one place in a very short time. Hence, it identifies more photons from spot B that spot A. Since we generate the interpolated image from "what is left" after clustering, B appears brighter than A in Fig 7.
If you have further questions or my explanation is not clear, please don't hesitate to ask.
Thanks!
------------------------------
csamuel
------------------------------
Original Message:
Sent: 07-14-2021 15:27
From: James Demarest
Subject: EDFA Magazine August Issue
Felix,
Perhaps you, the author, or Franco can help me understand this data. If I follow the discussion correctly the application of machine learning switched the region of interest of this analysis in intensity from the spot at (20,10) to the spot at (30,20)? I'm wondering if the authors could expand upon why this occurred and what it fundamentally means in simplified terminology... Thanks!
------------------------------
IBM
Albany NY
jjdemar@us.ibm.com
Original Message:
Sent: 07-12-2021 13:01
From: Felix Beaudoin
Subject: EDFA Magazine August Issue
Dear colleagues, the latest EDFA Magazine digital issue is available at the link below. You will find a guest column from ISTFA2021 Keynote speaker Dr. Ravi Mahajan (Intel), 4 technical articles, and an update on the 3 FA technology roadmap councils from Keth Serrels (NXP).
https://static.asminternational.org/EDFA/202107/
Please reply to this post if you have any questions or comments on any content in this August issue.
------------------------------
Felix Beaudoin, FASM
PMTS Yield Engineer
Global Foundries
Malta NY
------------------------------