BSC Computing vs BSC Maths and computer science

[tonycaw]

Hi guys. I am currently studying BSC computing part time via UNISA. I started last year(2020) and finished 6, 1st year subjects.
My aim is to be able to work in machine learning/AI fields.

I am in the position to chose between continuing in BSC Computing or changing to BCS Maths and computer science (see attached BSC Computing subjects)

I need advice on the following:

1. Is the math in the BSC Computing course enough to pursue a career in machine learning ?
2. Would it be more beneficial to have more math subjects?

 

[cguy]

The more more maths the better. Something else to consider though is that it is mostly applied maths and statistics that is the most relevant.

You want to ideally have linear algebra, differential equations, numerical methods, probability, distributions, multi variate linear regressions, stochastic methods, calculus, advanced calculus, perhaps signal analysis and convex optimization too if available.

I personally did a CS and pure maths major, two years of applied maths and one year of stats for my undergrad. I had to catch up a bit on the stats when in industry.

 

[tonycaw]

The more more maths the better. Something else to consider though is that it is mostly applied maths and statistics that is the most relevant.

You want to ideally have linear algebra, differential equations, numerical methods, probability, distributions, multi variate linear regressions, stochastic methods, calculus, advanced calculus, perhaps signal analysis and convex optimization too if available.

I personally did a CS and pure maths major, two years of applied maths and one year of stats for my undergrad. I had to catch up a bit on the stats when in industry.

Thank you very much for that detailed answer. Much appreciated

 

[LegionnaireJ]

Did vanilla CompSci through UNISA, choosing Machine Learning and AI modules for honours.
Would listen to @cguy on this one.

I would say the Math in the pure CompSci course is more logic oriented, and not what you'd ideally need in a working environment if AI/ML is the way you would like to go. If I could do it again, would do the Math + CompSci as you are considering. (If that math leans towards Applied Math).

My 2c regarding the math-y modules I can still recognise from your screenshots - its been a few years:

STA1610 - The stats module in the normal CompSci degree is fairly broad but shallow (being introductory). You will need to build way beyond this in terms of Stats.
MAT1503 - Linear Algebra - you will learn the basics about matrices etc. Ultimately also very light.
COS2601 - Some set theory / state machines.
COS3701 - Set theory continued, Lemma's etc.
COS3751 - Intro to AI. Very interesting, but very old as well. Good theory sections though( for what theory it does cover).
And then the formal logic modules. Loved it, but not necessarily all that useful in a ML career.

You are seriously making me want to go back and grab some Applied Math modules - insanely interesting.
Good luck.

 

[tonycaw]

Did vanilla CompSci through UNISA, choosing Machine Learning and AI modules for honours.
Would listen to @cguy on this one.

I would say the Math in the pure CompSci course is more logic oriented, and not what you'd ideally need in a working environment if AI/ML is the way you would like to go. If I could do it again, would do the Math + CompSci as you are considering. (If that math leans towards Applied Math).

My 2c regarding the math-y modules I can still recognise from your screenshots - its been a few years:

STA1610 - The stats module in the normal CompSci degree is fairly broad but shallow (being introductory). You will need to build way beyond this in terms of Stats.
MAT1503 - Linear Algebra - you will learn the basics about matrices etc. Ultimately also very light.
COS2601 - Some set theory / state machines.
COS3701 - Set theory continued, Lemma's etc.
COS3751 - Intro to AI. Very interesting, but very old as well. Good theory sections though( for what theory it does cover).
And then the formal logic modules. Loved it, but not necessarily all that useful in a ML career.

You are seriously making me want to go back and grab some Applied Math modules - insanely interesting.
Good luck.

Excellent. Good 2 also have input from someone who has done this course also. Never too late to grab the books again.
Thanks for your input.

 

[NR7]

Given how absolutely horrible UNISA has been with applications/changing qualifications and the fact that they are completely overwhelmed during this lockdown time (can only go to the campus if you book one day in advance for serious issues), I would stick with the BSc Computing since you're already 6 modules in.

 

[Messugga]

My view would be that it's better to hit the math hard during undergrad. If you're going ML/AI, you'll be doing a post-grad qualification. There are options during the latter for upskilling on ML but less so on math and stats fundamentals.
You may yourself struggling with some of the advanced topics without the appropriate foundational skills in algebra, probabilities, etc.

 

[tonycaw]

Given how absolutely horrible UNISA has been with applications/changing qualifications and the fact that they are completely overwhelmed during this lockdown time (can only go to the campus if you book one day in advance for serious issues), I would stick with the BSc Computing since you're already 6 modules in.

Yes also the subjects has changed from semester to year subjects. So i still have to figure that out.

But the first year subjects are almost the same. Everything I have done is also done in the maths/applied maths stream. For now I will choose subjects which are being done in both streams. Which gives me time till next year to battle.

 

[Ntobekoo]

Did vanilla CompSci through UNISA, choosing Machine Learning and AI modules for honours.
Would listen to @cguy on this one.

I would say the Math in the pure CompSci course is more logic oriented, and not what you'd ideally need in a working environment if AI/ML is the way you would like to go. If I could do it again, would do the Math + CompSci as you are considering. (If that math leans towards Applied Math).

My 2c regarding the math-y modules I can still recognise from your screenshots - its been a few years:

STA1610 - The stats module in the normal CompSci degree is fairly broad but shallow (being introductory). You will need to build way beyond this in terms of Stats.
MAT1503 - Linear Algebra - you will learn the basics about matrices etc. Ultimately also very light.
COS2601 - Some set theory / state machines.
COS3701 - Set theory continued, Lemma's etc.
COS3751 - Intro to AI. Very interesting, but very old as well. Good theory sections though( for what theory it does cover).
And then the formal logic modules. Loved it, but not necessarily all that useful in a ML career.

You are seriously making me want to go back and grab some Applied Math modules - insanely interesting.
Good luck.

Formal Logic may seem useless, but it helps with reasoning in general. Formal logic is useful in all disciplines remember Logic helps with reasoning. Logic comes before Mathematics. Logic helps to develop analytical and critical mindset.