NuclearPhynance Archive

Hi Jslade,



Like I said, I can field physics questions. I sent you an email with a few examples of ANN is particle physics. If other people are interested, here is an example:



MLP Higgs



The first interview I had talked about ML a lot, so that probably shades my understanding.

darkmatters -



I have no special insights to dispense here.



Instead, I simply wanted to you to re-examine your percepts. Media hype notwithstanding, most of the money in this industry is still made and lost using arithmetics. What you find more often than not in finance is that highfalutin mathematics is used as either a fig leaf or a bait.



Also, I have a question regarding that Higgs boson NN. Don't NN require a training set containing "true data"? How does one obtain such training set before discovering Higgs?

Temnik,



Check your email, I sent you a longer response.



We can calculate the higgs physics processes, then turning it into actual things measured into the detector involves using some previously measured physics.  All these are input into a Monte Carlo that can simulate signal and background, that can then be put into the ANN to train it.

Question about HFT: for somebody who's been working with exotic derivatives (structured credit and IR exotics), does moving into HFT constitute a danger of losing some of the maths skills? Are things which one learns in HFT transferable at all outside HFT domain?

Apologies if this is a little harsh, but that "T" in HFT stands for Trading. It's not high-frequency structuring, or high frequency stochastic calculus, or high-frequency exotic pricing.



If you're worried about losing your math skills, you probably shouldn't be going into trading. If you can find a way to use your math skills to make money, then you won't lose them. If you can't find a way to use them to make money, then losing them isn't a significant loss.



Of course, you can always do math stuff outside of work too, if that's your cup of tea.

"Apologies if this is a little harsh, but that "T" in HFT stands for Trading. It's not high-frequency structuring, or high frequency stochastic calculus, or high-frequency exotic pricing."



Point taken, but trading is the raison d'etre of all quant work, isn't it? Maybe my question should be rephrased: how much less do high-frequency traders need quantitative models than exotic option traders?



"If you're worried about losing your math skills, you probably shouldn't be going into trading. If you can find a way to use your math skills to make money, then you won't lose them. If you can't find a way to use them to make money, then losing them isn't a significant loss. "



I'm not worried about losing my math skills (just like I'm not worried about my IQ suddenly dropping to 80). I am worried about my re-employability as a quant after working in HFT.



Or maybe one should distinguish between HFT and algotrading?

Don't know how relevant this is, but one of the head honchos of research at knight trading gave several talks recently about using stochastic programming in HFT, that math is quite similar to path dependent option pricing.



Also, I wouldn't classify the signal processing type work as "less quantitative". Less focus on pdes, but similar issues tend to come up.

Would the algotrading team in Goldman Sachs a good place to work with such nice things?

Astar

isn't the reason that they gave the talk because it isn't useful to trading, but it does show that they are clever.