|January 26th, 2017, 02:18 AM||#1|
Joined: Jan 2017
Choosing between deep learning and numerical analysis
First I apologize for my bad English, it is not my first language Also I wonder if questions like this can be asked. If not I'll remove it asap.
I'm attending graduate school, just passed qualifying exam. From now on I have to choose specifically what to study. My advisor professor majored numerical analysis but now is interested in deep learning, and actually is working on it.
On my lab, I got to choose one of four topics : Deep learning, Computational Fluid Dynamics, Image processing, numerical analysis.
First, I don't want to do Deep learning. It is really hot topic now, and many people are working on it. It would be five years later that I graduate, and there would be so many experts. More over I really liked analysis, algebra etc. But deep learning seems less mathematical.
But what I'm worrying is that, Deep learning is really powerful so that other topics could be useless. Like I heard actually facial recognition can be easily done with deep learning.
So here's what I want to ask. Which way should I go? some numerical things or deep learning?
|January 26th, 2017, 06:13 AM||#3|
Joined: Apr 2014
Math Focus: Physics, mathematical modelling, numerical and computational solutions
I agree with Joppy; just pick the one that sounds the most interesting and go for it. Here's my opinion on those subjects specifically... maybe it will help you make a decision.
Deep learning: many algorithms and mathematical problems which were previously solved using conventional Monte Carlo statistics are now being solved by neural networks and genetic algorithms. These are going to be more and more prominent as AI starts to take a more active role in problem solving. It certainly feels like a "buzzword" topic, but there's a lot of potential mathematics to learn that will be useful for many applications.
Computational Fluid Dynamics: Even subsets of CFD are entire research fields in themselves because fluid problems are so interesting, challenging and mind-bending. If you're really interested in physics, meteorology or any other areas which involve fluid theory, you're going to love this because CFD has some of the most interesting and crazy-hard maths/physics/computing theories! Many CFD problems are solved using numerical computation techniques and you'll probably get a heavy dose of that too.
Image processing: a friend of mine did a PhD in this and it had some serious computing involved. I think there are machine learning techniques involved with these, so you'll get a bit of "deep learning" work here too... I think the only difference is the application.
Numerical analysis: all of the above topics involve numerical solutions to difficult maths problems, so numerical analysis gives you the tools to assess why those methods are used, which ones are most effective and why. You'll most likely learn about Von Neumann stability analysis, rates of convergence and complexity as well as a myriad of different methods for solving problems. Numerical analysis is very important and is often overlooked.
Regardless of which one you pick, you're going to get a massive dose of numerical computation work which will be valuable. If you're worried about the latter, it's worth noting that the company I work for has at least one expert in all of those fields to make sure we can we do anything that gets thrown our way I imagine most software companies do the same!
Last edited by Benit13; January 26th, 2017 at 06:22 AM.
|analysis, choosing, deep, deep-learning, learning, numerical, numerical analysis|
|Thread||Thread Starter||Forum||Replies||Last Post|
|Numerical Analysis||Faith||Real Analysis||1||July 7th, 2014 04:33 AM|
|SOS NUMERICAL ANALYSIS||evinda||Applied Math||9||November 3rd, 2013 09:22 AM|
|Analysis of Image superimposing using numerical analysis||saugata bose||Computer Science||0||March 8th, 2013 12:03 PM|
|Learning analysis?||forcesofodin||Academic Guidance||1||February 11th, 2010 09:31 PM|