Machine Learning

idkmyusername

Heya

So I just finished my degree and I have a major interest in Machine Learning,
I did some ML projects and got high grades in all of them (>70%, I hope that's good!)
The thing is I'm not really interested in Data Science... I'm okay with the amount of maths in programming but I feel Data Science is more maths than I can handle.
I was just wondering does anyone have any advice for me? How can I apply my interest in ML? Or should I give up altogether and do something different??

Thanks in advance!

srsly78

It's pretty much all maths, all this talk of "learning" and "intelligence" is overblown - it all comes down to applied statistics. If you can't handle the maths then it's not for you.

idkmyusername

srsly78 wrote: »

It's pretty much all maths, all this talk of "learning" and "intelligence" is overblown - it all comes down to applied statistics. If you can't handle the maths then it's not for you.

My ML projects weren't very maths centered, it was just programming algorithms. If there were statistics involved (usually something I would do myself to explore the data) it was still not that bad. From what I read, maths is mostly present in theoretical ML which isn't what I want to do anyway.

srsly78

Well I suppose it comes down to what you think of as "maths". Programming and algorithms are all branches of mathematics, but not really the scary type.

Did you encounter basic techniques like this? Most software engineers don't. However mechanical/electrical engineers probably are quite familiar.
https://en.wikipedia.org/wiki/Principal_component_analysis
https://en.wikipedia.org/wiki/K-means_clustering

I assure you that these statistical techniques are common in practical machine learning.

14ned

srsly78 wrote: »

I assure you that these statistical techniques are common in practical machine learning.

I can't emphasise this enough as well. Maths departments have been struggling to find good statisticians recently, between big tech and big finance anyone who can wave a stat is being gobbled up. For some weird reason if you're great at maths then either you're great at stats or you really are not. I guess it takes a certain type of brain to be really good at stats. It's not necessarily a maths brain.

(Most of us trained in stats can do it okay without a deep intuition for it. Some have a deep intuition for it, can feel out patterns instead of calculating them. They tend to make a lot of money, far more than I ever will)

Niall

idkmyusername

srsly78 wrote: »

Well I suppose it comes down to what you think of as "maths". Programming and algorithms are all branches of mathematics, but not really the scary type.

Did you encounter basic techniques like this? Most software engineers don't. However mechanical/electrical engineers probably are quite familiar.



I assure you that these statistical techniques are common in practical machine learning.

Absolutely, both techniques - as you say - are quite important and definitely have their uses. I guess the difference between studying them in a computing class and a statistics class was that the statistics involved no computers and was only theory and the whole maths behind it, whereas in computing it was simply applying these methodologies. I guess I would agree with you that the level of maths required in computing just wasn't scary

14ned wrote: »

I can't emphasise this enough as well. Maths departments have been struggling to find good statisticians recently, between big tech and big finance anyone who can wave a stat is being gobbled up. For some weird reason if you're great at maths then either you're great at stats or you really are not. I guess it takes a certain type of brain to be really good at stats. It's not necessarily a maths brain.

(Most of us trained in stats can do it okay without a deep intuition for it. Some have a deep intuition for it, can feel out patterns instead of calculating them. They tend to make a lot of money, far more than I ever will)

Niall

Absolutely, strangely enough most of my friends who studied Maths consider statistics to be a Social Science rather than a Mathematical field and they would have quite an aversion to Statistics...as well as any other Social Science!

Personally I definitely think even when doing Machine Learning projects, that a statistical exploration of the data, as well as, an understanding of the applications being used is necessary. Even when using scikit, after all it may need optimization and debugging.

I can visualize and interpret patterns, but I suck at doing all of the calculations (with the exception of regression it seems) or explaining the whole theory behind it!

srsly78

Are your friends still in college? No mature engineer/scientist thinks stats is a social science. Ever done a course in statistical thermodynamics? Not boring at all!! Neutron stars and exotic states of matter can all be explained via stats. Ever written a montecarlo simulation? All stats.

I will admit that I was also not enthused by statistics in the past, but have changed my mind after seeing the many applications.