Some random stuff here.
My name (宛Wan营Ying) has the same pronunciation as a common name for girls, 婉莹 (for example), I don’t mind being mistaken, but I prefer He/him/his.
As for the meaning of these names, I’m no expert in the Chinese language, but 婉莹 is roughly related to gentle and crystal-clear, while 宛营 abstractly means the hope of someone being strong both mentally and physically. And 宛 is also the city where my mom was born, so my full name gets something from both of my parent’s sides.
My non-work engineering interest is reading biological science books, I remember when I was little I was particularly interested in fossils (like Ross Geller), then later some documentaries, and now I’m just curious about animal cognition.
My interest in such topics turned out to be related to my Ph.D. work - the first work I did in my Ph.D. is Evolutive Network Architectures, and later such habit actually inspired a line we wrote in one of our explainability papers: “… within our current framework, it is impossible to fully isolate classifier behaviour from attack artefacts, since the latter are always learned using a specific classifier.”
My understanding is, the explanation we receive from any feature attribution methods shows (where these methods believe) where the classifier believes the artefacts are located. Consequently, each classifier acts as an eye to observe the physical / artefactual world. While some earlier types of eyes can only sense if there’s light or not, some modern eyes are fancy enough to tell purple from blue. When we report the observed artefacts from any classifier at hand, it’s like choosing an eye between, or beyond these two examples. Our goal is to accurately perceive the ground truth of artefacts; however, in practice, we cannot achieve that - the artefacts we observe always represent only a fraction of the entire ground truth. The world observed through the chosen classifier may seem color-blind when compared to a fancier classifier, or overwhelming when compared to a primitive one.
I do enjoy paper writing, it’s fascinating to me to organize findings and present them in a structured way, like building LEGO. And I think such interest is connected to my other interests in storytelling:
I also enjoy scrolling world map and reading Wikipedia articles of random cities.
| 1 | 8 | 9 | 5 |
“I reset that counter last night. This blog had nearly two-thousand hits in the last eight hours. This is your living …“ by Dr. John Watson in A Scandal in Belgravia.