2 ms·

Computer science isn't something that is becoming increasingly more mathematical, but quite the opposite way: computer science was seen as just another branch o

by jordigh 14y ago

Computer science isn't something that is becoming increasingly more mathematical, but quite the opposite way: computer science was seen as just another branch of mathematics. Alan Turing didn't quite have physical machines at his disposal to develop a theory of computation on, so he invented an abstract machine. Von Neumann thought that the foundational trouble in mathematics with Gödel meant that pure mathematics would soon be a sterile field, so he went on to design the Von Neumman architecture of computing machines we still basically use now. Read some Knuth or Dijkstra. They're mathematicians first, specifically, computer scientists. It was only in the 80's or so that computer science started to be seen as its own thing instead of a branch of mathematics. It's around this time that university departments separate from the mathematics department started to be created. When modern "CS" graduates come out of university thinking that an "algorithm" is some irrelevant theoretical tool that is only useful for implementing standard libraries of programming languages and forever forgotten, I weep a little.

graycat 14y ago
Didn't know the von Neumann story about his influence from poor Kurt. For 'more mathematical', really should say from when! Right, not from von Neumann! If pick when carefully, if only from yesterday, then since then the field has become more mathematical! Or pick from the start of Stanford Professor Ng's class on 'machine learning' which seems to borrow a lot from maximum likelihood estimation in statistics and steepest ascent optimization from mathematical programming or operations research. Or, CLRS -- they went to mathematical programming and borrowed linear programming. Once at IBM's Watson lab I saw some guys attacking load leveling via some carefully done work in stochastic optimal control. Design of distributed computer systems has long been one of the better applications of queuing theory. Once I needed to find nearest neighbors in R^n so started by doing essentially binary search on each axis (yes, after that need to do some tree backtracking with some cutting planes). I heard that there was such a thing, k-D trees, and, yup, found it in Sedgewick and later discovered that such work is called 'computational geometry'. And, for my last example (drum roll), finally there is some interest in the computer and network monitoring community in regarding monitoring as some continually applied statistical hypothesis tests, maybe! We will see if KPCB sends a copy of my paper to Endgame! In the short run, to attack a field with math can be one heck of a publishing opportunity. So in the long run, if only from 'academic competitiveness', a field doesn't have much hope as 'science' but to 'mathematize'.

by jordigh 14y ago

graycat 14y ago

Didn't know the von Neumann story about his influence from poor Kurt. For 'more mathematical', really should say from when! Right, not from von Neumann! If pick when carefully, if only from yesterday, then since then the field has become more mathematical! Or pick from the start of Stanford Professor Ng's class on 'machine learning' which seems to borrow a lot from maximum likelihood estimation in statistics and steepest ascent optimization from mathematical programming or operations research. Or, CLRS -- they went to mathematical programming and borrowed linear programming. Once at IBM's Watson lab I saw some guys attacking load leveling via some carefully done work in stochastic optimal control. Design of distributed computer systems has long been one of the better applications of queuing theory. Once I needed to find nearest neighbors in R^n so started by doing essentially binary search on each axis (yes, after that need to do some tree backtracking with some cutting planes). I heard that there was such a thing, k-D trees, and, yup, found it in Sedgewick and later discovered that such work is called 'computational geometry'. And, for my last example (drum roll), finally there is some interest in the computer and network monitoring community in regarding monitoring as some continually applied statistical hypothesis tests, maybe! We will see if KPCB sends a copy of my paper to Endgame! In the short run, to attack a field with math can be one heck of a publishing opportunity. So in the long run, if only from 'academic competitiveness', a field doesn't have much hope as 'science' but to 'mathematize'.