Should I take PDE before MFE?

[Mensa]

Hi QuantNet Community,

I am now a Columbia student, considering study plan next semester. I am preparing to apply MFEs for 2016 fall.

I am wondering whether knowing PDE well is compulsory or helpful for MFE program, or more generally for quantitative finance career? What other knowledge do you think I am still lacking?

I am taking: mathematical finance, programming for quant finance, stochastic models and data mining this semester. I plan to take numerical methods and stochastic calculus next semester. I am still considering what other courses to take.

Plus, I have taken the following relevant pre-courses in undergraduate:
1. Math: Mathematical Analysis I & II (cover Calculus, real analysis, ODE, etc), Advanced Linear Algebra, Probability Theory;
2. Stats: Intro to Statistics, Econometrics, Time Series Analysis;
3. CS: JAVA, C++, Database, Data Structures, etc.
4. Finance background is good, because it's my second major.

Any advice will be greatly appreciated!

 

[Daniel Duffy]

It does no harm to learn fundamental numerical analysis A-Z.

PDE is a bit more advanced. I don't know which MFEs ask it is used. Here are some theses several of which are PDE-based.

Datasim Financial Forums • View topic - MSc theses in computational finance

 

[Mensa]

It does no harm to learn fundamental numerical analysis A-Z.

PDE is a bit more advanced. I don't know which MFEs ask it is used. Here are some theses several of which are PDE-based.

Datasim Financial Forums • View topic - MSc theses in computational finance

Thanks for your reply. And sure, I will give a go for numerical methods/analysis.

Actually, MFEs never ask PDE as a pre-requisite, but it seems many quant finance models are dealing with PDEs. That's why I consider to learn.

 

[Daniel Duffy]

Thanks for your reply. And sure, I will give a go for numerical methods/analysis.

Actually, MFEs never ask PDE as a pre-requisite, but it seems many quant finance models are dealing with PDEs. That's why I consider to learn.

You're welcome
Many C++ quant library have a PDE engine and building your numerical knowledge incrementally will ease the transition.

A good model problem to learn is 1-factor heat equation using Crank-Nicolson and take it from there. If you understand the steps then you are well on your way to PDE Nirvana
https://en.wikipedia.org/wiki/Crank–Nicolson_method

 

[rajanS]

You're welcome
Many C++ quant library have a PDE engine and building your numerical knowledge incrementally will ease the transition.

A good model problem to learn is 1-factor heat equation using Crank-Nicolson and take it from there. If you understand the steps then you are well on your way to PDE Nirvana
https://en.wikipedia.org/wiki/Crank–Nicolson_method

rutgers university requires pdes and odes are prerequesite before joining the program

 

[rajanS]

You can take all the courses that you want mensa. In reality is, no one is going to hire you to do these in an actual job without a phd. And the jobs that require you to do that pde modeling is boring as fuck and you will be as far away from the money as possible doing that crap. I would mostly focus on practical math like data mining, statistical inference, regression, machine learning etc.

 

[rajanS]

You can take all the courses that you want mensa. In reality is, no one is going to hire you to do these in an actual job without a phd. And the jobs that require you to do that pde modeling is boring as fuck and you will be as far away from the money as possible doing that crap. I would mostly focus on practical math like data mining, statistical inference, regression, machine learning and programming etc where you can sell yourself.

An mfe is totally fucking useless. I have one. Way too overrated. Doesn't even help you land a interview. Most prefer people from a physics background. Just get a ms in cs or math from a top school and ur fine. An mfe is just a way to make money from u while teaching u useless real world skills

 

[Daniel Duffy]

You can take all the courses that you want mensa. In reality is, no one is going to hire you to do these in an actual job without a phd. And the jobs that require you to do that pde modeling is boring as fuck and you will be as far away from the money as possible doing that crap. I would mostly focus on practical math like data mining, statistical inference, regression, machine learning etc.

That's a shame. There's much much more to PDE than what you have ben exposed to. You probably had a teacher who did not like/know PDE?

Without PDE we would be living in the Stone Age, technologically speaking.

 

[rajanS]

That's a shame. There's much much more to PDE than what you have ben exposed to. You probably had a teacher who did not like/know PDE?

Without PDE we would be living in the Stone Age, technologically speaking.

Show me a job in the market that requires it. I went to jp Morgan, ms, credit Suisse, and looked up jobs. There isn't a single job that lists that as a skill needed

 

[rajanS]

Show me a job in the market that requires it. I went to jp Morgan, ms, credit Suisse, and looked up jobs. There isn't a single job that lists that as a skill needed

I have used crank Nicholson to solve finite difference methods in my numerical class. Bravo. Put that in ur resume and let's see if it gets u past the recruiters trash can

 

[rajanS]

Last, but not the least, the ratings online regarding the two books on finite differences and c++ pricing is enough to indicate how other quants view the book. Sure, some might find it as a mantra spoken straight from heaven. For others, it's nothing to shout on top of rooftops about.

If you want to go into pricing, by all means. Learn pde. No phd? Good luck in getting that job.

In a nutshell, the candidate asked if knowing pde is helpful prior to joining mfe. Sure it is

Is knowing pde helpful in getting a job? Hell no

 

[Daniel Duffy]

I have used crank Nicholson to solve finite difference methods in my numerical class. Bravo. Put that in ur resume and let's see if it gets u past the recruiters trash can

Minor point; it's Crank-Nicolson. Besides, it's a not very good method, see article here
http://wwwf.imperial.ac.uk/~ajacquie/IC_Num_Methods/IC_Num_Methods_Docs/Literature/DuffyCN.pdf

 

[ThanhBinh]

What is the "PDE"? I'm a newbie. Sorry if my question is stupid.

 

[rajanS]

Minor point; it's Crank-Nicolson. Besides, it's a not very good method, see article here
http://wwwf.imperial.ac.uk/~ajacquie/IC_Num_Methods/IC_Num_Methods_Docs/Literature/DuffyCN.pdf

I have never read your book. But if this is how you have written it, I look forward to reading it and improving my knowledge.

Thank you

 

[Daniel Duffy]

What is the "PDE"? I'm a newbie. Sorry if my question is stupid.

PDE is partial differential equation. Here is a relevant one for Black Scholes.

 

[ThanhBinh]

PDE is partial differential equation. Here is a relevant one for Black Scholes.

Thank you very much. This document is useful for me. I'm a computer science engineer. I'm learning about MFE by myself.

 

[Daniel Duffy]

Thank you very much. This document is useful for me. I'm a computer science engineer. I'm learning about MFE by myself.

You're welcome. A nice project to start with is Crank Nicolson for the heat equation. Once you have that then moving to Black Scholes is not difficult.

https://en.wikipedia.org/wiki/Crank–Nicolson_method

 

[Le Van]

Show me a job in the market that requires it. I went to jp Morgan, ms, credit Suisse, and looked up jobs. There isn't a single job that lists that as a skill needed

Perhaps you have overlooked it. Good knowledge of PDE modeling and/or Monte Carlo simulation is a must for some quant positions at Citi the last time I checked. Granted, machine learning and big data analytics are gaining popularity but that doesn't undermine the importance of PDE. At the end of the day it all boils down to where you'd like to work; buy sides firm generally want to hire statisticians and data scientists while sell side firms prefer mathematicians.

 

[rajanS]

Perhaps you have overlooked it. Good knowledge of PDE modeling and/or Monte Carlo simulation is a must for some quant positions at Citi the last time I checked. Granted, machine learning and big data analytics are gaining popularity but that doesn't undermine the importance of PDE. At the end of the day it all boils down to where you'd like to work; buy sides firm generally want to hire statisticians and data scientists while sell side firms prefer mathematicians.

Good point. I will try to improve my Monte Carlo/pde knowledge better. I have been focusing mostly on statistics and econometrics the past year. So I think I underestimated the importance of those subjects.

 

[Daniel Duffy]

Good point. I will try to improve my Monte Carlo/pde knowledge better. I have been focusing mostly on statistics and econometrics the past year. So I think I underestimated the importance of those subjects.

When possible, PDE is faster. But for more factors MC is needed.
Wilmott Forums - Numerical Methods Forum