Do all the things like ++ or -- rants, post your own rants, comment on others' rants and build your customized dev avatarSign Up
skprog20592yHo ho ho. Marry consumer consumption holiday #3
Meh, I'd be pretty hard to replace. I've not met many people that can code and do maths. Most of them work for the same company as me. Unfortunately I've interviewed many that claim they can. The hard part for me is replacing THEM. Not that many companies doing interesting R&D.
Root567402yTo be fair, it would take them several months and they'd find someone about half as qualified.
But the spirit of the message is true enough. 🙁
@alwaysmpe I can, I'm coming for that job muahahaha
@alwaysmpe joke aside, do you have any recommendations for working up to advanced data structures and algorithms? Don't get me wrong I have experience in it, but The DSAA book I'm reading focuses more on the concepts and gives you end of chapter programming challenges which is nice, but I want to get into the gritty math heavy tomes later on... currently self studying in Calculus. Then it will be discrete mathematics and lastly linear algebra. Is this a good line up math wise? I see you have a strong math background and I am trying to take that route as well, but self study it (not getting a degree in cs). I was under the impression that calculus and linear algebra (with statistics) would set me up nicely if I wanted to go the machine learning route and discrete mathematics would set me up for dynamic programming. Basically I want to open as many doors as possible... Thanks in advance!
asgs68162yBest piece of advice for long-term workaholics
@dalastTomCruise I'm happy to give my 2 cents but I can only speak for myself.
My background: I have an undergraduate degree in Physics and a postgraduate degree in Computer Science.
What we do is mainly motion tracking. We don't really use any machine learning techniques as we can do it faster without.
I have to say, having a bit of experience in both DSP and ML you need to know more to use DSP techniques for new problems but you get a much more concrete approach to specific problems. By concrete I mean it's possible to reason that the approach will be at least mildly successful. I've kinda come to view a dev saying "we should use ML" in a similar way to how devs view people that have ideas for products and are willing to give a share of the profits if you make it. They don't really understand what they're talking about. If someone can be concrete about what ML technique to use and why that's different.
That aside, dynamic programming and ML are hugely broadly applicable from motion detection , stock market analysis and bipedal robots to NLP, engine performance and electrical grid management. Unfortunately that means going as broad as possible. In terms of maths. I've been through K A Stroud, Engineering Mathematics, am reading the second book, advanced Engineering Mathematics and plan to read Steven Smith, DSP for engineers. Other things that I want to brush up on as it's ended up being relevant are graph theory and computer vision techniques. If you know what type of job you want to end up in try to get a broad background there too. That said when I started I was advised to read up about broadcast technology as that is the industry that we sell to. Unfortunately I used very little of what I learnt there, I should have been reading up on computer vision instead. I have some knowledge in those areas but I think I would have benefitted from more.
Work at the moment involves guiding our intern doing some R&D which has lead to implementing a specialised edge detector which is showing significant potential. One of our projects is a low cost low resource hardware implementation of existing IP and our existing approach to grouping objects is just too slow. We all have varying mathematical backgrounds with limited formal experience in computer vision, which makes me wonder what we missed, what else could be done better or faster using knowledge we don't have.