Exactly, I just have started a course of ML from Top down approach from you jump in ML with built in libraries and then slowly slowly you reveals the math behind that.
The math isn't really difficult (albeit SVD & linear algebra was challenging for me). For calc, the bare minimum is gradients, which isn't too difficult either. Though linear transformations that alter the dimensionality can make it confusing. Honestly, if you want to be really good at what you do, just learn the math. You'll be able to read research and understand things much more intuitively. There's even the opportunity for you to come up with novel optimizations or algorithms.
100% agreed - diving deep into the math during my undergrad & Master's really served me well. My POV is that beginners shouldn't lose sleep over it, but it's definitely important to learn
Thank god I'm not alone. Linear algebra was challenging for me as well but many of my colleagues have espoused about how it was one of their favorite math courses and wasn't too hard for them.
yeah cool and then when you want to be valuable at job market against people who read research papers with crazy math notation you need to know it and meaning behind it, basic math like derivatives is for begginers but in the end you need decent knowledge on statistics, analysis and linear algebra. Because guess who is gonna implement new methods quicker based on those papers? e.g. kaggle competitors need to know math.
@@lukamoz That’s a 100% valid point. In the video, I made sure to emphasize that these guidelines are for those starting out, but eventually they should dive deeper into the theory.
This holds true for engineering jobs. Can't deep dive into the ultimate goal of the field without maths. Looking for a well-paid jobs? Learn maths partly. You can't craft anything serious if you "just know maths".
Hey man I've been watching a few of your videos and I'm heavily inspired! I'm currently in Grade 12 and set to graduate high school in June 2025 and then take a gap year for my uni admissions and stuff. So during this time what do you think I should do to get a head start on ML? I love Math, programming and challenges. I'd love to hear your advice.
Ya I remember learning how to make a nerual network in Godot for my use from scratch by learning activation functions and gradient descent and learn more once I built. Thought this video was about ml isn't maths
This method will mostly pidgenhole people into using specific methods. This is fine if the field was some static non changing landscape, but it is not. Also, it would be giving people hammers and suggesting everything is a nail.
@@nomadbl That’s a valid point. I assume you’re referring to Classical ML methods vs Deep Learning methods? Classical ML is still 100% relevant in 2024, even if it doesn’t get the same hype that DL does. My recommendation for those starting out is to first cover topics like Linear & Logistic Regression, and then move on to Feedforward Networks. They should come back to SVM, etc at some point for sure.
@@gptLearningHub I personally came into the field from the DL perspective and learned to appreciate the classical statistics literature (still learning!) as well as older ML methods. So I know the dangers so to speak. I think the approach you outlined can work as a starting point, just making sure that students are aware of the statistics literature as well as the range of ML tools and their intended purposes.
Yup, I have a 100% free course here with over 10 hours of lectures: www.gptlearninghub.ai/full-llms-course I also offer a $9.99 course for those looking to get started ASAP: www.gptlearninghub.ai/first-principles-framework
Thank you for sharing your view. My view is that math is still 100% important for ML, but the basics of calculus and linear algebra are all that's needed to get started
Check out my video on Gradient Descent - the explanation only requires understanding single variable calculus: ruclips.net/video/bbYdqd6wemI/видео.html
More Resources!
Math Review for ML: ruclips.net/video/OYJhBjnLp0I/видео.html
Gradient Descent Explained: ruclips.net/video/bbYdqd6wemI/видео.html
Linear Regression Review: ruclips.net/video/2vE3DqWdEXo/видео.html
Neural Networks Explained: ruclips.net/video/xZcOTAJ-h6w/видео.html
First-Principles Framework (Learn Fundamentals): bit.ly/40Kktiq
Beginner's Blueprint (Build Projects): bit.ly/3Z7qYdV
Im so happy to be part of this community. I love 3Blue1Brown videos and you're basically him but for ML
Wow, thank you so much! His animations and explanations are definitely in a different league, but I appreciate the support regardless :)
Exactly, I just have started a course of ML from Top down approach from you jump in ML with built in libraries and then slowly slowly you reveals the math behind that.
Glad you found this insightful :)
The math isn't really difficult (albeit SVD & linear algebra was challenging for me). For calc, the bare minimum is gradients, which isn't too difficult either. Though linear transformations that alter the dimensionality can make it confusing.
Honestly, if you want to be really good at what you do, just learn the math. You'll be able to read research and understand things much more intuitively. There's even the opportunity for you to come up with novel optimizations or algorithms.
100% agreed - diving deep into the math during my undergrad & Master's really served me well. My POV is that beginners shouldn't lose sleep over it, but it's definitely important to learn
Thank god I'm not alone. Linear algebra was challenging for me as well but many of my colleagues have espoused about how it was one of their favorite math courses and wasn't too hard for them.
yeah cool and then when you want to be valuable at job market against people who read research papers with crazy math notation you need to know it and meaning behind it, basic math like derivatives is for begginers but in the end you need decent knowledge on statistics, analysis and linear algebra. Because guess who is gonna implement new methods quicker based on those papers? e.g. kaggle competitors need to know math.
@@lukamoz That’s a 100% valid point. In the video, I made sure to emphasize that these guidelines are for those starting out, but eventually they should dive deeper into the theory.
This holds true for engineering jobs. Can't deep dive into the ultimate goal of the field without maths. Looking for a well-paid jobs? Learn maths partly. You can't craft anything serious if you "just know maths".
I'm gonna pretend i know what you are talking about.
1)At least know single calculus thing
2)Click links
3)linear regression
4) you are ready for ML.
Hey man I've been watching a few of your videos and I'm heavily inspired! I'm currently in Grade 12 and set to graduate high school in June 2025 and then take a gap year for my uni admissions and stuff. So during this time what do you think I should do to get a head start on ML? I love Math, programming and challenges. I'd love to hear your advice.
Ya I remember learning how to make a nerual network in Godot for my use from scratch by learning activation functions and gradient descent and learn more once I built. Thought this video was about ml isn't maths
Yup the title is definitely a bit clickbait haha. Maths and Gradient Descent are super important for ML, in my opinion.
good videos keep it up
Thanks for the support man
This method will mostly pidgenhole people into using specific methods. This is fine if the field was some static non changing landscape, but it is not. Also, it would be giving people hammers and suggesting everything is a nail.
@@nomadbl That’s a valid point. I assume you’re referring to Classical ML methods vs Deep Learning methods? Classical ML is still 100% relevant in 2024, even if it doesn’t get the same hype that DL does.
My recommendation for those starting out is to first cover topics like Linear & Logistic Regression, and then move on to Feedforward Networks. They should come back to SVM, etc at some point for sure.
@@gptLearningHub
I personally came into the field from the DL perspective and learned to appreciate the classical statistics literature (still learning!) as well as older ML methods. So I know the dangers so to speak.
I think the approach you outlined can work as a starting point, just making sure that students are aware of the statistics literature as well as the range of ML tools and their intended purposes.
@@nomadbl Definitely. Prob/stat and its various applications can't be ignored. Thank you for sharing your experience!
Sorry I'm a bit confuse do you offer any free course or anything because I only see web that have no link with a 500$ price tag?
Yup, I have a 100% free course here with over 10 hours of lectures: www.gptlearninghub.ai/full-llms-course
I also offer a $9.99 course for those looking to get started ASAP: www.gptlearninghub.ai/first-principles-framework
🔥🔥🔥
I appreciate the support!
machine learning it is math
100% agree - at some point, everyone should dive into details like backpropagation - I just think it's not the best introductory topic
@@gptLearningHub But ML is math
No need to learn math because we have AI for it
Highly disagree, I do not understand a thing in ML because I don't have any background in prerequisite maths.
Thank you for sharing your view. My view is that math is still 100% important for ML, but the basics of calculus and linear algebra are all that's needed to get started
Check out my video on Gradient Descent - the explanation only requires understanding single variable calculus: ruclips.net/video/bbYdqd6wemI/видео.html