Machine learning and AI is extremely easy if you learn the math: My rant.

[@ChemCoder404]

Some useful books in the description. HERE's the playlist you all requested: kzbin.info/aero/PLRfh8v6mu_F1tC9pX3zML149ddL0TxdCV Share your experience here. How important was mathematics in your learning journey? kzbin.info/www/bejne/g3ene6enq7N5Y6M

[@eliasmai6170]

machine learning involves functional analysis for optimization issues, and computational topology for topological data analysis

[@jaredfontaine2002]

I love the linear regression etc we learned it in business school to find out stock beta then I found out that it was machine learning...

[@mdebarshi]

Mathematics is the most natural thing... and ideally, it shouldn't be hard... Mathematics is the only thing that is fully driven by logical thinking, where one (correct) step automatically leads to another... Communication of "mayhematical" ideas is indeed hard... and it is the "teaching" (verbal communication of mathematical ideas) that makes mathematics unnecessarily hard. Learning mathematics by thinking (done by yourself) is easier than someone else doing the thinking for you and then converting their thoughts into words while "teaching". Instead, just spend aome time with yourself, think and teach yourself. (It is not necessarily that the speed of your brain has to match your teacher's brain speed or the speed of their speech... you're "slow", could mean, you think deeper and in a divergent (exploratory) manner and hence you "lag" behind.) I found mathematics easy when i spent time on myself and learnt things by my own... at my own pace... ______________________ Abstraction should be banned while teaching or learning mathematics. You can do all the abstraction once you have mastered a topic, until then, things should be extremely granular (and simple). Where there is no clarity, there is no understanding. But of course, we can always take some initial help into a topic and not reinvent the wheel each time... but it is crucial that we don't build a civilization without the wheel. The wheel must be present at all costs and should never be sacrificed to abstraction.

[@Dawsatek22]

before i code i always try to learn some math

[@paulvmunix]

I have a BS Physics and a BS Computer Science from 85-92. Mathematics has been the critical element for my entire career. Tensors FTW! AbeBooks is imho the best and most economical source for textbooks. I coded my first inference engines on a Jetson Nano. I quickly found that to progress I needed to bone up on matrix and tensor math. Excellent channel Sir!

[@fireinthehole2272]

Hi, ML PhD here: learn the math. It might not be super fun but it will be worth it and it gives you tools you will use your entire career.

[@Reminder100]

I'm in high school rn, for the past month I've been doing mix ML projects no math and learning math, do you this approach is good or should I just learn all the math then start doing projects, thanks! 🙃

[@ChemCoder404]

I would still say start with the math. Earlier the better. Have you had any linear algebra classes? Like matrix operations etc.

[@Reminder100]

​@ChemCoder404 well I live in quebec canada (10th grade), it's complicated to explain but in simple terms I can't take those classes, I'm slowly studying more advanced things compared to what they teach

[@ChemCoder404]

Awesome! So good that you started early. What made you interested in this field?

[@jasondaniels640]

How you still enjoy learning after AI takes over..

[@d.youtubr]

Necessary roadmap (books) for a noob: 1. Elementary algebra for school - Knight 2. Higher algebra - Knight 3. Calculus made easy - Silvannos 4. Problems in calculus of one variable - Maron 5. First course in Probability - Ross 6. Introductory statistics - Ross 7. Linear Algebra - Strang 8. Practical Statistics 9. Basic Multivariate calculus

[@ChemCoder404]

Awesome! Thank you for sharing this.

[@saptarshisanyal6738]

Excellent roadmap, but this step will take 2 years to complete.

[@TerriTerriHotSauce]

Or just use Khan academy for free with interactive exercises

[@ansar107]

I don't think I can do this while studying my bachelors

[@ajamosci4901]

Please what's a realistic time frame to learn this Math and Statistics?

[@dominicellis1867]

Gaussian curves, logistic curves, langrange multipliers for gradient descent, basic group theory to understand backpropogation, lots of linear algebra for the weights and bias, and regression analysis to train your model. It’s all basic undergrad math.

[@ChemCoder404]

Yup. But it’s not taught everywhere or isn’t a requirement until people want to do some ML.

[@jackvial5591]

What group theory do you need to know for backprop? I think you can develop a good enough understanding by learning gradients, chain rule, and gradient descent

[@ajamosci4901]

@dominicellis1867 Please what's a realistic time frame to learn this Math and Statistics?

[@dominicellis1867]

@@ajamosci4901 it depends on where you are in your math education?

[@sifodyas_]

How does group theory apply to backpropagation?

[@Vishnu-n9l]

Yes, please teach us machine learning with a focus on the mathematics behind it, starting with perceptrons and vectors. Comparing machine learning with mathematics will help us learn more quickly and effectively. I am eagerly waiting for your upcoming video.

[@ChemCoder404]

Thank you for the support. Yes, it’s on the list. May not be the very next video but I hope you stay tuned. 😊

[@trifalgarh]

Doesn't Andrew Ng's course already do this?

[@Vishnu-n9l]

@@trifalgarh Don't know, That course is free or cost? If it is free could you share the link

[@doofus8]

​@@Vishnu-n9lnot possible to share links on yt as it tends to block them... just google ml specialization by andrew ng & you'd get to coursera page where you can take the course for free

[@Protract_Loop]

​@@ChemCoder404ok

[@katalyst4stem]

Hi ChemCoder! I’m a doctor, specializing in diabetology, and I spend quite a bit of time analyzing patient data (good old Excel). I’m really excited to dive into your videos to refresh my math skills, especially as they relate to machine learning. Thanks for all your hard work-more power to you!

[@ChemCoder404]

Thank you so much for the encouragement! It’s certainly on my list. It may not be the very next video but I hope you stay tuned.

[@katalyst4stem]

​@@ChemCoder404 Subscribed and 🔔 notification pressed, Sir 😊

[@ChemCoder404]

Thank you so much!

[@ronbackal]

Hi Katalyst! I am very curious - what kind of data are you analyzing? My father sadly passed away from Diabetes. Thanks, Ron

[@katalyst4stem]

@@ronbackal Hi Ron, I am sorry to hear that. there are many parameters that I look at. Chiefly Cardiometabolic Health: HsCRP, IL-6, Apo B, Direct LDL levels and Lipoprotein A aka Lp(a) 24 hour - ambulatory BP measurement In select cases I ask for doppler of neck (Carotid-intime media thickness) Renal: urine protein to creatinine ratio, Cystatin C and serum creatine / GFR Musculoskeletal Health: Bone density (forearm and ankle) Body composition analysis (using InBody to measure total fat%, visceral fat %, muscle % and lean body mass) Pulmonary Health: Pulmonary Function Test and Breath Holding Test Eye Health (remember eye is the window to the brain): Fundoscopy looking for retinal changes from diabetes and high BP You can visit my page and check out the videos under Heart Diseases explained (it will lead you to my 2nd channel .... lose2win club >> to help 10,000 ppl lose 1 lakh kg body weight and analyse the impact on overall health) I hope, I have answered your question ...

[@ayushjha3716]

Fun fact: The math is hard

[@darylallen2485]

😢

[@praveen921]

I think it depends what level you're going for. For most people who are working on business use cases and not on the cutting edge of technology, a solid understanding of calculus, linear algebra, statistics, probability, and maybe some basic differential equations is enough I think. I'm from India though, and unfortunately, these foundations aren't stressed enough. People dive right into libraries without the intuition.

[@nitishvidyadharan8240]

I finding learning language, arts etc tough as compared to maths. Why dont people talk about it. I am finding also my son is facing the same problem. He is relaxed during its maths exam but get very tensed in subjects like history, hindi, sanskrit etc.

[@ravirajchilka]

Nope

[@jasondaniels640]

Fun fact: it's easy for AI

[@edwardgrigoryan3982]

Thanks for this video. It provides validation for the path I have chosen for myself to eventually learn ML/Data Science. I have dedicated 1.5-2 years just to learn the math (clearly I'm not in a big hurry and this won't apply to everyone). I just finished reviewing precalc, and am going to move on to doing the entire calculus series and linear algebra, along with a major emphasis on stats and eventually the ML specific math that is needed to be highly competent in the field. Currently work as a product manager/data analyst for a startup, but want to gradually upskill to ML/Data Science over the course of a few years. It just didn't make sense to me to skip or rush through the math. Perhaps it helps that I actually enjoy the math quite a bit. Anyway, once again, thanks!

[@ChemCoder404]

Thank you for sharing that and for your support!

[@dencentbeatz794]

Keep going man that’s dope af. Any advice on how u structure ur learning after ur outta college. I also just wanna keep learning more math stuff as tech develops. Still in school right now, but I mean I’m taking all the ML classes so it’s way more structured to learn the math.

[@edwardgrigoryan3982]

@@dencentbeatz794 Hey, thanks man. It's cool that you're this interested in this stuff while so early in your journey. You have PLENTY of time to learn and explore :) Honestly, what I do is very simple. I think it's mostly about the discipline and consistency than it is about medium or method, though of course that matters too, just way less. I just make sure I study some math at least one hour a day, ideally more if I can, for 5-6 days/week. The weeks and months add up. I use a combination of lecture series on KZbin, some paid courses from The Great Courses (those are okay tbh, not nearly deep enough but alright for big picture context and overview), and textbooks. I honestly think I should be using more textbooks, as it really makes you work hard and think, so I'm going to be trying that more for the next few months. One very important thing: I try to have a balance of theory/concept AND what I call operation/procedure. It's easy to do too much of one or the other. I'm trying to understand the theories and concepts, essentially the deeper "why" and "how" and I work hard at that, but I also make sure to do plenty of problems to be able efficient at working through them and making sure I can apply the concepts well and under all sorts of different circumstances. Honestly, there's so much out there, just try a bunch of learning/studying formats and methods and you'll find what works best for you. Then, stick to that with a reasonable level of discipline and consistency. Wish you the best in your pursuits.

[@ChanceTEK]

Yes, I am extremely interested in learning the math behind Machine learning. I am a Data Engineer transitioning to become a Data Scientist. Becoming more involved in Generative AI. Thank you.

[@WahabKhaddim]

yessss ! finalllllly someone talking about this I thought I was the only one who noticed that no one teaches maths... Pleaseeeee make a playlist I beg you !

[@ChemCoder404]

It’s coming up. Stay tuned. Thanks for your support!

[@basuutube]

i am interested in learning the math behind. will eagerly await for your math playlist

[@ChemCoder404]

Definitely on my list. Thank you for the support

[@LeFrog0]

I'm actually teaching an introduction to supervised deep learning for undergrad programmers this year. It's supposed to be a maths course, so I've decided to deconstruct the pipeline and focus on simple examples, explaining the maths. The first 3 or 4 lessons just focus on non-convex optimisation in one variable, gradient descent for a linear regression, with or without stochasticity, and the PyTorch framework using `backward` and an `Optimizer`. Then I'll introduce MLPs and transformers. I've only had one lesson so far, but the students have been really receptive. Their comments and your video are encouraging me that I'm on the right track, thanks!

[@ChemCoder404]

Awesome. Thanks for sharing and the encouragement.

[@chickugangavarapu772]

damn thats great . can i get access of the resources if thats fine

[@Roarzambimaru]

You’ve pinpointed what I feel has been missing in my understanding of ML: The math. Would totally watch if you made more math-focused ML content!

[@ChemCoder404]

Coming up soon

[@samsonv9332]

Yes pls make a video with math concepts behind ML, thank you 🙏🏼

[@kirillholt2329]

I've learned math on my own time (had a background already) but what still escapes me is the motivation behind it. Why some machine learning algorithms are the way that they are.

[@LostinMango]

What kinda maths you learned ?

[@kirillholt2329]

@@LostinMango linear algebra, multivariable analysis, probability and a bit of stats.

[@AnA-xx1vx]

If you didn't develop intuition, you didn't learn the math. I've decent knowledge of math but same issue ig , probably because knowledge is there not understanding,which perhaps comes from spending time with something.

[@doofus8]

​@@AnA-xx1vxI'd personally recommend "mathematics for machine learning" book to you.... it's so fundamental & relates everything with machine learning that it's inevitable for you to not build the necessary intuition for ml algorithms

[@AnA-xx1vx]

@@doofus8 I've read mml , probably I should try again. Actually major problem is lack of clarity of goal setting. Thank you for your suggestion.

[@Camstraction]

I completely agree with you and I wish that there was more of an emphasis on the math in machine learning courses. It is very understated how important it is to understand the math behind the algorithms in order to apply machine learning models in the real world, where often the use cases are not cookie cutter like presented in courses. The only way to appropriately choose the right algorithm, loss function etc.. is to understand the math behind it. I mean even our evaluation metrics matter there are many limitations that have to be understood and it all lies within what the math actually tells us about our data and6 model performance. I am a math major and enjoy pure math for its own sake but there is a real (and big) gap between the math and it’s implications in applied settings. It’s not an easy gap to bridge and often requires jumping back and forth (without good direction) between your specific application and the seemingly disconnected math. This is because the way math is taught doesn’t lend itself well to generalization in the real world by nature of how math is taught but also by how math is made. I often find myself trying to find resources to bridge this gap and like you’ve mentioned, it’s hard to find any. It really would be a great help to myself and many others for people like yourself to help bridge that gap!

[@ChemCoder404]

Totally agree. Thanks for sharing this! I hope to bridge that gap too.

[@raisingthesteam]

You're right, bridging that gap is key, especially for those with a strong math background. Dr. Stephen Odaibo, who was a guest in episode 8, wrote *The Foundational Mathematics of Artificial Intelligence* to connect math with AI concepts. It could be exactly what you’re looking for! Available on Amazon.

[@Endou47]

I'm 100% interested in hearing more about your takes/maybe even some classes or book recomendatations for foundamentals of math for ML. Please keep it up

[@ChemCoder404]

Thank you so much! Next video on resources and recommendations.

[@ckjdinnj]

The linear algebra and calculus is arguably the easy part. It’s the discrete math and implementation details of applying the techniques. The best videos I have seen is tsoding’s series. He doesn’t give the best in depth math but he walks through implementation from scratch in c with no 3rd party libraries. The insight gained from going through an implementation makes it so much easier to gain an intuition for the math.

[@ckjdinnj]

I would say any video that starts off with python and installing a library is worthless. It’s impossible to understand what the parameters for the library functions mean without understanding the implementation.

[@reallynotpc]

When I started learning maths, machine learning hadn't yet been invented! But, yes, I did do some basic statistics and I think linear regression was part of my A-level syllabus. That's a good rant you have there! Happily, I am rather old and will never need ML to make a living.

[@ChemCoder404]

Thank you for sharing that!

[@MalamIbnMalam]

In my graduate school program for Masters of Computer Science they expected us to already have the mathematical foundation in Calculus, Linear Algebra and Statistics/Probability. So it was not reviewed. I had to review it on my own.

[@ChemCoder404]

Yeah graduate school is usually like that.

[@MalamIbnMalam]

@@ChemCoder404 yep, you are right

[@Iamine1981]

Coming from a Maths background, I found ML as very intuitive. A lot of the algorithms boil down to simple processes such as: building a design matrix X with your training/test data and features, pre-processing this data, picking up an algorithm to train on the data depending on the task at hand, MAYBE defining a custom loss function if the default one is not appropriate? And finally, some version of SGD or some other optimisation procedure for learning the parameters. It is not much more complicated than that for basic ML. Areas to focus on are: Matrices and Linear Algebra, statistics and probabilities/distributions, and a few algorithms for numerical optimisation.

[@ChemCoder404]

Summed up nicely! Thanks for sharing!

[@StarDynamics]

Someone finally had to make this video. Thank you I've been watching numerous deep learning videos and reading many DL books since 2019. I didn't knew any math beyond simple arithmetic. throughout all these years i never got any further with AI. I was a hobbyist . I never understood why I don't get the concepts and why I can't properly visualize what they are talkin about. Recently I realized it's not my lack of intelligence that stopped my progress. I simply didn't know any math and aaaaaaaaaaaaalll those videos and books skipped the math. I started learning math from algebra and am learning more and more every day. It really amazes me how understanding simple math concepts like a slope of a line and the Y intercept can be more useful for understanding machine learning compared to several courses and books. Nowadays I'm more interested in the math itself than machine learning. constant failures in understanding deep learning made me feel like sth is missing and i started learning math and i'm loving it. I'm in my 30's and considering going back to school to study engineerinng. That's all because all those shitty tutorials that skipped math made me feel I don't need it. but the truth is, that is the only thing I needed to understand anything in engineering.

[@ChemCoder404]

That's awesome! Thank you for sharing!

[@kakabudi]

Right now in my undergrad a mandatory class to take is statistical foundations for data science and machine learning. So far I’ll say it has very little to do with what people’s basic concept of statistics is. A lot of it is linear algebra and multi variable calculus. In class we do problems by hand/on paper. For hw we write the program in Python and document EVERYTHING about the process on an Overleaf doc. Tldr: the math is somewhat involved. Reports take a long time. Hopefully by going through the conceptual math process you’ll have a great understanding on the inner workings of ai and ml.

[@ChemCoder404]

This is fabulous!

[@freeideas]

I understand a bit about matrix multiplication, but so does numpy. I understand a bit about how the weights and biases are used between the layers, but so does tensorflow. I understand a bit about using calculus derivatives to find the minimum loss, but so does keras. I make neural networks all the time, but I'm not sure how I would use any math directly.

[@ChemCoder404]

I think your level of understanding is good enough.

[@Iamine1981]

An underappreciated aspect of learning the maths behind ML, at least the supervised branch of it, is that it shapes your brain to start looking into datasets mostly as design matrices X with m numbers of instances, and N features whether those are categorical features, ordinal, real values etc. This will help tremendously when you are thinking about feature design/engineering, normalisation, feature selection algorithms, etc…

[@ErroneousTheory]

I’ll see you and raise: what is the solution architecture? What processes are running and what do they do? How are they configured and how do they communicate? Memory management? Ipc? This is just a computer program, so how do the programs run on computers - all the way down to the copper? No one seems to know.

[@Iamine1981]

Took me 2 YEARS of assiduously going through the most common algorithms and procedures for Supervised Learning only, and these are standard algorithms when it comes to ML. The math behind them is captivating and simple enough, and this made the learning curve very enjoyable to be honest. I am in agreement with you that a LOT of "data scientists" out there do not have a solid enough grasp of the math powering the algorithms in popular packages such as Scikit-Learn, and anyone who spends time going through the hard math will be beautifully rewarded later, as ML - at least in its most basic form - is not that complicated if your math skills are good.

[@ChemCoder404]

I agree.

[@rolimiranda9291]

A few months ago 1 got a book about neural networks that explains that back propagation alogorithm and kind of caught my attention , so what I did is I learn just enough calculus/partial differentials and some matrix operations (Im a busines major so I kind of have to learn from scratch) ,a few days after,I tried to do the literal number crunching on pieces of paper with just a simple neural network, all that I can say is it's amazing!

[@ChemCoder404]

Yup. I’m going to show another way that will help in the next video. Stay tuned!

[@Thumper_boiii_baby]

Please do a full video on ML math ❤

[@ChemCoder404]

It will probably be a playlist. But I hear you.

[@phil97n]

I haven't taken any machine learning courses but I'm learning machine learning on my own through books and videos. For some reasons I intuitively figured that math would be that important for machine learning so I spent at least a year and half to two years just learning math on my own: linear algebra, calculus, probability, statistics. At least 7 books in total - smart folks can do with less. I was very excited to learn machine learning but kept resisting the temptations to start before being comfortable with math. I think machine learning just makes more sense when you understand the underlying math theories and algorithms

[@ChemCoder404]

Yup. Thank you for sharing that!

[@sarak8467]

Yes, we need the ML Math playlist please.

[@ChemCoder404]

It’s coming up.

[@brendawilliams8062]

No. Thankyou for the chance to say no

[@Bllackdog]

I feel like the actual rant is supposed to be around the fact that you actually don't need the math for it :D. I think many people that are trying to train models heavily rely on already built systems and libraries. They take validated datasets from online resources and tweak a few parameters. Then they continue with "Try and Error” for some time until they reach a point to be satisfied with. I don't know how many of them are really interested in analyzing own datasets in depth like what a high bias results in or a high variance. What it means, if grad. desc. oscillates with huge amplitudes. The libraries are enablers, which is kinda great I guess. At the end they won't wrap their mind arounds the actual architecture, since they cant really understand what's happening under the hood but build extensions for the architecture thus founding systems of use. At the end im pretty happy how I started my journey into this field by keeping distance to acutall NNs for several months, focusing on math. The math is the real fun regarding AI, NNs etc. For me that's the point where logic and creativity meets to make things happen.

[@ChemCoder404]

Thanks for sharing!

[@TheStudent-df5bc]

Yes please! We would love it!

[@dylanbyrne01]

I’m about to enrol in a MSc in AI. Mathematics is a HUGE part of the course. Linear algebra, statistics, calculus and probability are massive elements of the introduction to machine learning module, and take up a good chunk of my first semester.

[@ChemCoder404]

Awesome. Learn well.

[@dasikalyan]

Yes, the Introduction to Machine Learning course by Professor Andrew Ng does an excellent job of covering the foundational mathematics required for the course. Having been away from a college setting for more than 30 years, I appreciated how methodically he introduced the necessary mathematical concepts and explained the science behind various statistical methods, leading up to the different machine learning algorithms.

[@ChemCoder404]

Yup. I agree.

[@patrickchan2503]

at school we learnt all the maths for the sake of learning maths without a use case.

[@ChemCoder404]

I feel you. But math is almost everywhere now.

[@suzukigsxfa9683]

Coursera has a 3 part math certificate in linear algebra calculus and statistics. I started on this before picking up a book on Python coding for ML

[@ChemCoder404]

Awesome!

[@redowansakib592]

I finished the course and now reading a book on mathematics for machine learning. The course teaches the basics but I wonder would that be sufficient.

[@ChemCoder404]

To apply ML and problem solve, probably yes. But what if you are asked about it during interviews like I was?

[@deletedaxiom6057]

I recall, taking an ML class in university and when the professor wrote the formulas for back propagation on the board i was so confused as to how you took the derivative relative to W which was a fixed value at that step. I later ended up double majoring in math just because I couldn't stand not understanding all the math. I wanted some kind of intuition. I think of ML as there is some ideal shape to whatever knowledge you are training a NN on. The training data is like topigraphical samples of it. After enough samples and fitting the curve of your NN to them you have a basic outline of that idealized knowledge.

[@AlemMemić]

I started as a mathematician, with bachelor/master/phd all in theoretical math. During my phd I realized how my research area is applicable in Machine learning. That was in 2017. So, it was the other way around. First math, then Machine learning. I had difficulties in understanding how everything works when it comes to software engineering. Mathematical brain is different than programmers, so I had to switch.

[@ChemCoder404]

Interesting. Thanks for sharing!

[@Imisambi]

Yes Yes YES!!! Very interested in seeing the mathematics involved with machine Learning, deep leaning & Artificial Intelligence in general

[@ChemCoder404]

There’s a playlist for that! Check out the channel page. Yes, I'll be adding more to that playlist

[@hemanthdev8882]

For sure!! im looking forward to the videos

[@ChemCoder404]

Thank you. There’s a playlist on the channel. Check it out!

[@JustaSprigofMint]

After feeling lost for almost 10 years in my career I think I'm ready for a switch. AI has completely captured my attention and imagination and I really want to know how to get into this. Sadly I don't have a STEM background, rather from the media. What can I do?

[@ChemCoder404]

I help people get into AI and ML. So, you're in the right place! Here's a roadmap that can help you and stay tuned for more: kzbin.info/www/bejne/h3eoqYJnhsdpgtU

[@prajwaladhav3123]

Would love to see a series you can make on maths related to ML

[@ChemCoder404]

Here’s the growing playlist: Math for AI and ML kzbin.info/aero/PLRfh8v6mu_F1tC9pX3zML149ddL0TxdCV

[@Rsx2OO9]

I agree with this. As a AI researcher machine learning is basically all math. I was never that good at math before I started but the cool thing is if you become passionate about ai research and machine learning you eventually grasp the mathematical concepts behind a lot of the algorithms. But yes for someone new math geared at machine learning would for sure be a great foundation. and having a lot of libraries that simplify the process for example PyTorch are good tools for experimenting with concepts to help reinforce the mathematical ideas

[@ChemCoder404]

Well said.

[@babatundeonabajo]

I agree with this video. It reminds me as well of the people who want to learn React without first learning JavaScript.

[@Ghost____Rider]

A very good book is "Mathematics for Machine Learning" . Also yes I'd be super interested in a video or a playlist on the math behind ML!

[@ChemCoder404]

Thank you for the encouragement. It’s coming soon

[@ragtop63]

College and university degree programs in AI/ML require you to take math courses. It's only online that you see courses in these topics without math.

[@ChemCoder404]

Totally agree.

[@danirichard470]

A friend of mine got his MS in Computer Science with a focus on AI. I counted that he has 50 math books on the mathematics of Machine Learning. I am reviewing my Linear Algebra and Multivalent Calculus cause I too want to understand the mathematics and make ML work.

[@ChemCoder404]

That's awesome! Keep it up!

[@adrianguerra1952]

Of course, with all my heart. Learn the maths is life goal for me beyond AI, but ùath for machine learning is an extreme good start. Thanks !!!!!!

[@ChemCoder404]

That’s good to hear

[@ml_bangla]

Hey ChemCoder! I appreciate this video of yours. In today’s hype-driven world, people dive into Machine Learning without understanding the math, which is so crucial. There's a funny observation I’ve made about my university teachers-they routinely split the dataset into a 7:3 ratio for training and testing, then fit the model and check the accuracy. They don’t seem to understand the mathematics behind Machine Learning. I agree it can be challenging to balance learning the core concepts of Machine Learning with the underlying math, but it’s essential. Otherwise, the foundation would be weak

[@ChemCoder404]

That is a really good point.

[@fiohannshanahan-dover3416]

Interested in the math for sure! Would definitely study any content you put out on this

[@ChemCoder404]

Awesome. There’s a growing playlist on the channel!

[@Godsman_42]

Thank you for explaining that to a T. Yeah I’m actually picking up textbooks so I can learn the math. At 43, I think in about 1 or 2 years of studying the math, I should be ready to apply the other easier part.

[@ChemCoder404]

That's awesome! Keep going!

[@dencentbeatz794]

That’s honestly why I took the stats version of Machine Learning right now lol. Yea are problem sets are both theory proofs and then we have a computational part. Def hard lol but interesting. Gonna try and take the rest of the Stats ML undergrad and some grad courses if I can

[@MrHaggyy]

I think this really depends on what you are doing with machine learning. If your focus is creating content with the model, like a image generation website, it's ok to keep it a blackbox and exchange the model as needed. If you apply this to any important data, like corporal statistics, finance or engineering, you really need to know enough math so you can judge the capabilities of a model. And if you want to build and train models on your own, every bit of math in that subject you truly own is a real gamechanger on how confident you can work with AI. Especially the time you need to code them up goes down significantly, and yes calculating a good size and error margin is really helpfull.

[@ChemCoder404]

Thank you for sharing that!

[@tieronspear9606]

Yes! Teach us the math behind ML.

[@ChemCoder404]

In the works. Thank you for the encouragement!

[@Jennifer_63165]

I would be interested in learning the math behind ML!!!! I just found your channel! Thank you for sharing!

[@ChemCoder404]

Thank you! I’m working on a playlist specifically for the math.

[@ToluDara0234]

@@ChemCoder404 Just Subscribe because you're willing to teach the math. Looking forward to the playlist!

[@Antoniosoldittt]

I’ve taken 2 AI/ML bootcamps. Neither dove deep into math and one of my instructors couldn’t even pronounce ordinary least squared. However, I haven’t found it difficult to develop accurate linear regression or classification models for finance.

[@ChemCoder404]

Thanks for sharing. Not many folks want to dig into the math as it is harder to teach.

[@johnanderson290]

Dr. Andrew Ng’s series on Machine Learning, although a bit older, is quite good, and focuses heavily on the mathematics behind ML.

[@ChemCoder404]

Yup it does. I'm not sure if it's still there on Coursera.

[@lokesh4008]

Yes sir! I have been looking for math required for DS and ML. Plz make videos on Linear Algebra, Statistics and Calculus.

[@ChemCoder404]

Certainly on the list

[@agenticmark]

gradient descent understanding or lackof is not dangerous.... loss functions are not difficult. you do need a high level understanding but you dont need to be able to do DC in your head to build good models. overfitting isnt the problem it once was.

[@nicolasgonzalezl]

You should make a video highlighting the specific topics in mathematics that are most important for AI/ML. A tree showing the ideal order to learn these things in would be good.

[@ChemCoder404]

Absolutely on the cards!

[@wademackey1098]

I got my engineering degree on a nine year plan , going part time at night. As a resulat I took ALL of my math courses before I ever took and engineering course. Engineering was sooooo easy if you already had the math. Same thing when I dabbled in ML later in my carrer.

[@ChemCoder404]

Awesome!

[@AK-NikhilSuresh]

Did you made playlist for Maths for ML?

[@ChemCoder404]

That’s coming up next!

[@AK-NikhilSuresh]

Thanks Brother , waiting for Playlist ☺️

[@inafog]

By no means I am a good ML guy but i am still decent and learning , we have tech clubs in out college and i am in the ai team so i had to take some interview for people who wants to get into ai team , i asked multiple people how to transpose a matrix and people were not able to do it.

[@ChemCoder404]

That is concerning! Thanks for sharing.

[@skeed1496]

I decided I want to get into machine learning professionally about a month ago and started with learning Python. I’ve since completed a course and have a beginner-intermediate understanding. After this video, I think I’ll go ahead and start working on statistics until I’m ready to keep learning 😂

[@ChemCoder404]

That's awesome! Take a look at our Roadmap video on the channel to get some ideas!

[@jumpwer]

I am very lucky to have been a part of IIT Madras...they have taught us Maths really deep before teaching ML

[@ChemCoder404]

Awesome!

[@pratiknapit8494]

The basics of LA, calc and stats are not enough as well. It takes deep knowledge of both pure and applied maths to understand deeper concepts in ML/DL. For example, if you don't have a strong background in real analysis or topology you won't understand the ideas of manifold learning or why proving RNNs would be better than normal MLPs for sequential data. Even the notation in advanced DL textbooks are so hard to understand without at least an undergrad level of maths.

[@ChemCoder404]

Thanks for sharing! 🫶

[@Ewakaa]

Trust me. LA, Calculus, Stats and Probability. Trust me you are good to go. Like really good to go. That’s why a Bachelors Degree really helps.

[@Ewakaa]

The funny thing is. These topics are very deep. Like I spent 2 years Learning Calculus from Pre Calculus to Calculus III. But some Genius wants to learn it in 1 week. Yeah sure you can. But it’s really deep. Like Inhave a calculus book that’s over 1,000 pages😂

[@ChemCoder404]

Indeed. College helps for such basic level of math which is needed.

[@RM-lt8rg]

They zipped through the math and if you couldn't keep up, the attitude was "Oh well!". I personally couldn't keep up so I studied for hours every week, learning the foundational math (Linear Algebra and Multivariable Calculus), and then took a week off of work before the midterm, so that I could teach myself Artificial Neural Networks, and solidify my understanding of the other concepts. It was well worth the effort and I recommend to anyone starting out in ML to do the same. Kahn Academy was my source for the Math skills. I also did a Statistics course in Zybooks as well as a Python course in Zybooks. Those you could probably get elsewhere, but I used Zybooks. For the Math though, I would absolutely go with Kahn Academy. It is the best!

[@ChemCoder404]

Interesting. I have heard they’re good. Maybe I’ll check them out to see if I can add something here.

[@RM-lt8rg]

@@ChemCoder404 I've never found anything like them before. You start their Multivariable Calculus course and realize you need a refresher on derivatives, so you enroll in their AP Calculus course and start that program. Your then realize that you need a review of Trig, so you enroll in that. Then you come back and pick back up with AP Calc, then multivariable. It's like a recursive program returning with everything that you were looking for. And it's free, but we should donate to them because they are doing a serious service to humanity.

[@brimendis]

Can you give more detailed examples of the problems you'll encounter with ML models if you don't know the math vs knowing the math?

[@ChemCoder404]

You will encounter problems regardless. But knowing the theory, you can problem solve faster and be confident about the model you choose and the method you follow.

[@joecavanagh1297]

I’d like to offer a suggestion: it may not be the best approach to dive straight into the required math when first exploring machine learning. For someone who has just developed an interest or curiosity in the field, I would recommend starting with the aspects that initially caught their attention and experimenting with those. This way, they can gradually build on that interest without being overwhelmed by the more challenging sub-skills, like math, which might feel discouraging at the beginning. My advice would be to first become proficient in the areas that excite you, and then gradually develop mastery in other essential areas, such as mathematics, as you progress.

[@ChemCoder404]

Agree. Thank for sharing that!

[@Gughuagendjsywhebbbjduww]

@@ChemCoder404I am already sold on ML and have used some of it, so I like the idea of jumping into the math

[@chrisogonas]

Totally agree! Learning the related ML math is foundational and makes ML mastery very well grounded.

[@ChemCoder404]

Yup. How did you go about learning and getting into ML?

[@chrisogonas]

@@ChemCoder404 Thanks for asking. My research required me to apply ML, and so I took a deep dive into self-learning. So much of my ML journey has been self-learning and application. I also took a few courses at the university that could help me get more grounded, but they felt like a scratch on the surface. I keep exploring through self-learning.

[@hackbauer9297]

I just started an online course that starts with statistics and they strongly recommended getting the Springer Statistics books. Would like to see your take on the mathematics.

[@ChemCoder404]

Yes, I am making a video on it. Will be out soon.

[@ridhimarhythmj5954]

Could you list all the topics needed for ML AI?

[@ChemCoder404]

That’s coming up in the next video!

[@rawatmemer262]

Would Love a playlist ! ❤❤

[@ChemCoder404]

Check out this playlist, I’m adding more here: Math for AI and ML: This strategy makes learning so much easy! kzbin.info/www/bejne/pqrbk2WCqZJ-eqs

[@amortals508]

I would very much enjoy learning the mathematics required for ML, or any mathematics in general! If you decide to start teaching I’ll be watching!

[@ChemCoder404]

Yup. It's certainly on my list. Planning a playlist.

[@5imian]

Yes please do videos on the underlying math, I would subscribe for that

[@rikudouensof]

I did the maths in Computer engineering school days. Makes a lot of sense.

[@erinomani9105]

I haven't even started yet . Just browsing and trying to figure things out. Would be awesome to narrow down the math courses to learn . Subscribed

[@ChemCoder404]

Thank you for watching! Stay tuned for more.

[@muhilan8540]

there's a great book called mathematics for machine learning

[@erinomani9105]

@@muhilan8540 Thanks . Very much appreciated. Do you happen to remember the author?

[@kevinmcfarlane2752]

Foundations are Linear Algebra, Calculus, and Probability & Statistics. I would also bookmark 3Blue1Brown's Essence of Linear Algebra series. It's useful for consolidation, if you happen to have started somewhere else. I find that when learning new technical concepts I have to read and/or watch more than one presentation to get comfortable.

[@erinomani9105]

@@kevinmcfarlane2752 Thank you Kevin. Very much appreciated

[@asim-gandu-phenchod]

Would love to see your math topics related to ML

[@ChemCoder404]

Thanks! Check out the growing playlist on the channel: kzbin.info/aero/PLRfh8v6mu_F1tC9pX3zML149ddL0TxdCV

[@jackvial5591]

Highly recommend doing Karpathy’s Neural Networks Zero to Hero course here on YT, use this to focus your math learning in a top down approach. In school they teach you math from bottom up or try to, it’s very demotivating for most people, myself included. Khan Academy has videos on all of the multivariable calculus and statistics you will need to know, be selective about which videos you watch, you don’t need everything in the multibariable calculus course. Also learn newton’s method and the history of root finding algorithms, this is where gradient descent and other iterative optimization algorithms evolved from

[@ChemCoder404]

Thank you for sharing that!

[@artukikemty]

For advanced ML you need to master linear transformations, projections, and if you like go for wavelets, fourier, SVD decomposition

[@architech5940]

I took some free online courses through edx, and it was my experience that you should have a foundation in linear algebra and statistics as a prerequisite for machine learning (at least for the free courses). They do go over some basic linear algebra before getting started with regression models, but not anything rigorous nonetheless. After realizing that I should become good at linear algebra and statistics, I picked up a statistics book from the library. Luckily the book I picked up had a focus on data science (sort of).

[@ChemCoder404]

That’s great! Thank you for sharing.

[@yesno8204]

Bro thank you. No one even talks about this unless you go and look for it.

[@ChemCoder404]

Agreed.

[@whatyousaydere]

just taking a second to appreciate how crisp you camera quality is.. can see this good lookin' man spit bars in 4k without my glasses on

[@ChemCoder404]

lol thank you.

[@brettfine3444]

Hi, would you please teach the math subjects and topics required for ML? Thank you.

[@ChemCoder404]

Yes, I am planning a playlist. Stay tuned!

[@laternite]

I don’t see why anyone wouldn’t want to learn, understand, and how to use the math.

[@AZyzk]

Thanks for the video! I would like to know more about the math behind ML please.

[@ChemCoder404]

Coming soon!

[@77jcarva]

It would be really useful a further mathematical explanation behind models.

[@ChemCoder404]

Planning a playlist.

[@crocopie]

So what are the maths involved? Calculus, Linear Algebra, and Statistics?

[@ChemCoder404]

Mostly those three. But you only need a select few concepts.

[@dyahns]

Andrew Ng does show the math in his Coursera course, not in depth as it’s just a 2.5 month course, but all matrix operations and formula behind algos in the curriculum are presented and explained.

[@ChemCoder404]

Yup. That’s where I started.

[@Harmxn]

We were taught the math behind the algorithms, just to understand what the model actually does. We didn't go very in depth into it, but we got a good idea

[@ChemCoder404]

That’s good

[@danieltriana1937]

Yeah, I'm interested in learning the math behind ML, but a good start before teaching them by your self, would be to create a road-map video for study them on our own

[@ChemCoder404]

Yup and here’s a video I made just for that: Roadmap to Transition into AI and ML in just 6 months! kzbin.info/www/bejne/h3eoqYJnhsdpgtU

[@danieltriana1937]

@@ChemCoder404 I had already saved the video to my "Watch Later" before your reply. I didn't know, haha! Very usefull videos. Thank you!

[@ChemCoder404]

I’m glad. Thank you for watching!

[@Lolleka]

Physics PhD here. ML enthusiast. Learn the math, folks. It's where the fun is, really. If you are into it for the magic you should be willing to learn the spells.

[@ChemCoder404]

Haha yes

[@nubinshred]

Thanks for vocalizing the need for this

[@ChemCoder404]

Thank you for the support.

[@c.mirashi]

I've taken this project of SER that's Speech Emotion Recognition And either I've to do A research or a project on this as a web interface And I'm clueless Idk where to start and don't know what to learn and how much of it exactly Currently I'm reading research papers

[@ChemCoder404]

Find a mentor who’s done it. Interact with them!

[@calvinrob4068]

Loved this vid. It touched on a distinction made by the other GOAT Andrej Karpathy on learning and leisure. A lot of educational content is an epsilon away from leisure despite being marketed as education given that it has a larger target audience than true educational resources. He says that instead of taking the snack of educational leisure which create the false impression of overstated learning, we should pursue the meal of true learning. Similar anology to the difference between the 15 minute full body workout versus a solid gym session. Maybe worth checking out the original. This channel appears to be true learning and not educational leisure!

[@ChemCoder404]

Thank you so much!

[@haripurushothaman1310]

Yes. I would love to learn that part, sir.

[@ChemCoder404]

Coming up...

[@CodeBrewCS]

Yes lol in my ML and DL courses we learned all about the math begind everything! We learned about the math for forward, backprop, CNNs… you name it (but i did my courses in uni). We had to basically go through whole epochs by hand 😂

[@ChemCoder404]

That's great!

[@ΝΕΚΤΑΡΙΟΣΚΟΥΡΑΚΗΣ]

Great idea, math tutorial for machine learning!

[@luisoncpp]

4:45 imho learning matrix multiplication or determinants shouldn't be a part of a machine learning course, people should learn linear algebra before taking that course. For me it sounds as crazy as trying to take a machine learning course without knowing how to code. However, I'm ok with including gradient descent in a machine learning course rather than putting it as a pre-requisite, mostly because many tricks to make the gradient descent faster (like mini-batch or using relu activation functions) are specific from machine learning. The gradient descent itself is a very simple idea, so it's better to give it a review at least to agree in the notation. Even backpropagation itself is not very well known outside machine learning (numerical optimization's literature mostly gives for granted the computation of the gradient, and calculus courses teach how to compute a gradient by hand rather than a method practical to code in a computer), so I think teaching backpropagation is a must in a ML course.

[@ChemCoder404]

Thanks for that! I am planning on a video about gradient descent and backpropagation.

[@nr9058]

I really liked your Roadmap video as it's more detailed. Unlike a lot of other "roadmaps" out there, Your advice of being more hands on by practicing with datasets seem to make sense. According to my professor, The only way to get a hold of the math for machine learning is by implementing papers (will be super helpful if you can make a video about implementing a paper or maybe do a tutorial !) . I have been following your videos for quite some time and they have been very helpful. Tysm!

[@ChemCoder404]

Thank you! Implementing papers do help a lot. It’s on my list and hopefully I can make it soon.