In my *undergrad* computer science courses, we were taught things like how to prove an algorithm delivers some defined transformation of data structures within some time and space complexity bounds.
It would be cool if we aspired to that level of formalism as the standard for promulgating AI programs.
@Unampho it was fascinating to take machine learning theory which at least in my undergraduate program was mainly linear algebra++ and high level differentiation and then the actual projects which were “throwing shit against the wall until it works”
at risk of projecting my own undergraduate ignorance unto an entire field, definetly sometimes feels like ML is closer to something to Software Engineering (if it works it works) than other computer science academic fields.
@terminalgoblin @chluebi in university of Michigan, I was taught machine learning with more theory and properties of metrics that applied across different methods. The very same course number another semester taught software engineering neural networks.
My mobile robotics course was a nicer progression to the bleeding edge, going through derivations of SLAM methods (kalman and particle filter variants) onto things like ratSLAM and paper replication projects near the end.
No real point here, just saying.