I’ve been crawling through all the EVE related data and I came upon this dev blog from way back:
I noticed this is a problem ideally suited for some ML tinkering (https://en.wikipedia.org/wiki/Simulated_annealing + some clustering algorithm for example). Have any of you devs implemented anything along those lines yet or does this sound like a good approach?
Why apply overkill when it’s not necessary?
Feeds techie egos, kills companies but so what, looks awesome on resumes!
I find that in general where the answer is “use current in vogue buzz words” that the problem has been poorly understood and that the solution isn’t being driven by the question, but by the interests of the Answerer.
Machine Learning is about pattern detection and categorisation. It has several weakness. It matches based on reinforcement based on training data, with all the weaknesses inherent in that data set - critically biases and low probability targets. Also there is no determinable reason for a decision made by the system. You can’t ask “for what reason was this categorisation decision made for this item/case?” and that’s a real nightmare.
Yes, new technologies and ideas are interesting and exciting. But interesting and exciting doesn’t mean they are suddenly the right tool for a job. Having a wider range of knowledge and approaches to tackling problems is always better (and why experience matters).
Remember: If you only have a hammer then not everything is a nail.
This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.