Network Analysis thoughts

This paper was quite interesting, and informative, despite its short length it certainly was not topical and gave a good breadth to the fundamentals of what is becoming an increasingly important science today. The node theory has so many implementations from social media to crypto to nature systems it is real and has effects. There is a lot of talk about constructing framework for networks and then analyzing results or their movements. To quantify a network gives one crucial data to better interact, optimize or grow it.

There is one thing that I think is left out which is that data is not enough, you need something called intuition, and ability to regard forces outside of the nodes. What I mean by this is that to understand something you can not just give it a dataset and have it be quantified or train an AI on, there needs to be an outside touch. Networks don’t exist in a vacuum and unless you build multiple meta systems & lots of iterations you can get stuck in loops.

Case in point in WW2, the Allies we’re loosing lots of planes and with them valuable pilots. One can not put armor all over a plane because it will reduce its agility and speed. You must be strategic. Engineers started tracking bullet hoes from planes that returned from the war and plotting them on a model to get a better understanding of where planes needed that armor, as a increase in armor reduces bullet effectiveness.

Here is a good example of these charts. Those tasked with analyzing made the connection that Germans were shooting at edges of wings as that is where they are thinnest and easiest to get to tear, loosing a piece of the end of the wing greatly compromised control over aircraft and could send it into a spin. In addition the middle was an easy target and where munitions were stored so it made sense fire was concentrated there.

Efforts were made to add hull strength to the new aircraft off the assembly line of the plants. Yet it didn’t matter the rate of downed planes did not change. The pattern was the same but no difference. In a closed system we know bullets cause damage, damage increases likelyhood of a break, and additional hull strength mitigates that. Without and outside spark the conclusion was doomed to repeat itself.

It took someone outside the system to note, that the people on this project were being short sighted. Obviously the armor did nothing. The reason planes were being downed was because shots in the exact areas that returning planes did not have were in crucial parts of plane. IE: the crewmembers and gas reserves! It takes a external system to feed novel data into other systems to actually produce working models.

My point is to say when we talk about systems we must acknowledge that there is no closed loop. To have conclusions you must continually iterate at higher levels, if not given the correct parameters no amount of machine learning will solve this.