@tpg2114 Not that I really understand what's going on, but I guess that does kind of look amenable to machine learning to an extent. To sample the solution space somehow and thus make the optimizing faster. But I guess that's kind of what you were saying earlier.
@tpg2114 Is it more than just a combination; If the results per se are not new, is there new discussion or something to justify a new paper (say, something along the lines of a literature review)?
@NovaliumCompany Well what you're describing is a genetic algorithm, and I don't see why you couldn't apply that to neural networks usually: i.e. you create a bunch of neural networks by applying random noise to the previous generation in some way and then you check which ones do best and select those to generate the next generation. Normally though in neural networks you update the weights by back propagation (basically algorithmic differentiation)
@enumaris Ah ok, fine, though I think not only are universities educating a lot of data scientists, there's bootcamps and all to that effect, so I would argue the opposite and in my immediate surroundings see that data science job postings want people to hit the ground running, i.e. machine learning experience (either from a previous job or some academic credentials to that effect)
In academic physics the math can indeed be the hard part (though even there I'd say coming up with an interesting problem to solve is the most difficult thing), but not really in industry.
@NovaliumCompany Math is a creative subject. Would you be afraid that reading poetry somehow limited your creativity and out of the box thinking ability?
@SirCumference Well if you ever do end up applying for a software engineering job, the typical entry level interview is about intro algorithms, so it's not a waste of time
@SirCumference I guess the answer's going to depend on many things (eg what do you consider software dev - there's many specializations). I wouldn't be surprised if you were unable to get an interview at Big N if you didn't have CS/ML on your CV (be it as a minor or some projects to show interest in and some knowledge of the subjects).
@NovaliumCompany Even if your ambitions are entrepreneurial rather than scientific (indeed I guess you have chosen the wrong chatroom), college opens doors. Most importantly you'd be networking, and even if you say that the skills you'd learn in college would be close to useless, the piece of paper you get at the end is not worthless as other people do ascribe value to it. I guess the best example of a degree of that type (where extracurricular activity is particularly important) is an MBA.
@alex1stef2 Thijssen's book covers a lot of subjects. If you're interested in some more specific subject, then probably a book in that particular area would do best.
@ACuriousMind Well it depends I guess what the library does. Ideally the external APIs were relatively well designed and/or at least the library itself was not bloated with things it's not supposed to do. For example, your SDE solving library should be doing just that, and the piping of the associated metadata should be handled by a different system. Replacing the internal library, then, is less work for clients (though by no means is it ever going to be zero).
@KyleKanos I don't think I've ever seen that approach work (though it's often tried). I think the best chance of getting something done is to scrap the codebase and start afresh (sell it on building on newer paradigms - both codewise and mathwise), though I've seen and read about that leading to messes as well.
@KyleKanos Well I guess you could pimpl everything. But it sounds more like a company culture issue of course if styles cannot be enforced (lack of reviews or mandates mixed in with a legacy codebase etc).
@KyleKanos I found tools like include-what-you-use useful, but it's all a bit project and build dependent as to how much use such tools will be and how much customization they require of course
Probably all the journals on Beall's list send out invitations to publish/come speak at conferences to anyone whose name and email address they can get a hold of (i.e. basically anyone who has previously published anywhere). I used to get at least a couple a day, but my spam filters are better now (and I think people in general now know about predatory practices and so it's probably less lucrative).
@enumaris That's just what recruiters (or whoever's pay depends on you accepting the offer) want you to believe. Hiring's such a massive pain nobody with any sense is actually going to rescind the offer for asking more time if they find someone good for the role.
I mean, that makes sense if you think about how the buy side and sell side in principle operate, right; one "takes risk", the other "sells risk and hedges it away" in very crude terms.
Also, a quant doesn't mean a quant, either and Wikipedia will list several different kinds. You have the guys doing theorem - proof and stochastic calculus because they've decided that no-arbitrage is the framework to work in and that's then very mathematically rigorous; and then you have the guys doing algotrading where you're doing stats and trying to find signals.
@ZeroTheHero It all depends on what those terms exactly mean, right. I have friends working in, say, "market risk" at big banks and maths-wise what they need to know is the Taylor expansion (though I bet if you read the job description, it would sound like something very advanced).
And not least because there nowadays are degree programmes specifically tailored to those needs (similarly, I think the door is slowly closing on physicists wanting to get into data science in that there's now so many people who are specifically taught the methods and so a physicists general aptitude to pick up new stuff to compensate for lack of education is less valuable)
Also, working in the financial sector doesn't necessarily mean that you're one of the ones to do any mathematics there; I would bet the vast majority of physicists working in the financial sector are not doing anything related to any kind of quant finance.
@ZeroTheHero So I'm surely guilty of having occasionally been the physicist from the xkcd strip, but if doing thing X is your goal, I'd read literature written by people in X, rather than that written by physicists to fit their own background and experiences.
I do find it very surprising that the equations of physics can quantitatively describe nature with such a few variables (though this does kind of break when you start taking more interactions and try to describe them, like in stat mech etc.)
@ZeroTheHero So I guess the exact topic of discussion matters, but in quant finance the idea is to, first and foremost, write models that by construction exactly hit the market---only after achieving this would one look at its behaviour. This is quite different from physics, I think, where one tries to write out some equations and then see where they go and what kind of things they imply.
@ZeroTheHero These examples are I think closer related to economics than they are to finance (sure, some of the Nature Physics special issue looks to be related to networks of some kind in finance, but I'd say that's still a very economics-like take on the subject).
