How I’d learn ML in 2024 (if I could start over)
Become better at machine learning in 5 min/ week 👉🏻 newsletter.borismeinardus.com/
In this video, I share how I would learn Machine Learning in 2024 if I could start over.
For the past 3 years, I have been studying machine learning (and 2 years before that basic computer science), which has now led me to work with an amazing ex-Meta professor, collaborate with Google DeepMind researchers, and have interviews at amazing companies.
Having learned from all of my failures and successes, this video breaks down how I would learn machine learning all over again, focusing on the essentials and learning from the best resources.
Enjoy!
👉🏻 Book a one-on-one call
calendly.com/boris-meinardus/...
👇🏻 Links to resources 👇🏻
==== Maths ====
www.edx.org/learn/probability...
www.edx.org/learn/linear-alge...
www.coursera.org/learn/matrix...
==== ML/ DL ====
www.coursera.org/specializati...
• Neural Networks: Zero ...
www.coursera.org/specializati...
huggingface.co/learn/nlp-cour...
⬇️ Follow me on my other socials and feel free to DM questions! ⬇️
⚫⚪ Medium: / boris.meinardus
🐦 Twitter: / borismeinardus
================== Timestamps ================
00:00 - Intro
00:40 - Python
01:29 - Maths
02:47 - ML Developer Stack
04:00 - Learn Machine Learning
06:06 - How To Really Get Good
=============================================
#ai #learning #machinelearning
🚀 There is so much more to explore in ML. Feel free to grab my FREE cheat sheet of different ML domains and open challenges: boris-meinardus.ck.page/2f5e05bb1f
0:35 0:39 0:41
We need a video for how to get a job in ML?
It will help you. 1. Basics of python 2. Learn numpy, pandas, matplotlib 3. Beginner course: * Supervised Machine Learning: Regression and Classification * Advanced Learning Algorithms * Unsupervised Learning, Recommenders, Reinforcement Learning By- Andrew Ng 4. Neural network * Neural Networks: Zero to Hero 5. Deep learining specialisation * Neural Networks and Deep Learning * Improving Deep Neural Networks: Hyperparameter Tuning, Regularization and Optimization * Structuring Machine Learning Projects * Convolutional Neural Networks * Sequence Models
As an MSc Computational Physics student and a beginner in learning machine learning who has done some research on how to teach myself ML, a lot of what you said is consistent with my own conclusions and how I would approach the self-learning process. Great video! Subscribed.
As a PhD student in NLP, I completely approve your recommendations! These are relevant, concrete and feasible steps so thank you very much for presenting this in that very pleasant way. I would maybe add something regarding college students that may wonder if they should start with math or computer science majors to work in machine/deep learning top companies. My advice relying on my experience is: start with math. It is, to my mind, far easier to learn computer science concepts when you already got the maths principles. I do have a math background initially and I struggle a bit at the beginning of my journey with some software engineering aspects of project development, but I am convinced that I would have struggled way more if I had to learn math concepts from scratch by the end of my degree.
Thanks for sharing this!
Pin this comment already!
Thank you very much for the insight! I really appreciate it and value it, that you approve of what I say 💛 But yes, maths is always the biggest blocker for people wanting to get started with ML. I think both paths (Maths first, or maths second) are viable and depend on the person. I know some (probably most) people might be frustrated right away when they just start with pure maths (I personally love maths :)). That's why I think starting with something fun like programming basics is a good idea. As mentioned in the video, no step really needs to be fully completed one after the other. You can also learn maths and python side by side. Mix and match to your desire, as long as you put in the time, and enjoy it, everyone can learn machine learning. And If you are in college doing CS, you will have linear algebra and calculus courses anyway haha. You just need to appreciate the content and power through! Again, thank you for your comment! 😊
This video could have been a 3-4 sentence guide. Disliked
what do you mean by "math"? Do you mean more than second year stats, calc1-3 and linear algebra? if yes, then perhaps the math major recommendation makes sense.
Thank you for this! I'm starting my neuroscience PhD soon, and I want to implement ML to aid my projects. Your video provides a super helpful framework going forward :)
Fantastic video Boris! And excellent practical steps, especially the final one! (Courses = base knowledge, projects/paper replication = specific knowledge)
Many thanks! You summarised it spot on! 💛
Bist ne gute Seele! Danke für die Motivation.
underrated video, underrated channel . Gonna binge watch all the videos ! so informative and straight to the point, please make more videos on machine learning and artificial intelligence
wow, I really appreciate the kind words!! Thank you 😊 I will do my best to not disappoint you! Let‘s see if you like this weeks video 😬
This video was extremely time-effective and simple. Thanks for putting it out!
Happy to know I followed the right order in the first 4 steps! Thank you for recommending Andrej's course; I subscribed to Ng's Course but I couldn't keep up with backpropagation :( Will include this step in my learning journey!
Words are not enough to express my words , i was just too confused to even start even tho i do have prior basic knowledge, as someone coming from software engineering having a basic idea i did had hard time figuring out what to learn , i cant think just how confused how other people are . Thanks a lot 🔥❤
I am a Business Consultant with 2.3 years of experience, and I'm planning to transition into the Data Science field. I was just looking out for guidance on how to get started. Thank you so much! I found it very informative.
Great content. I started my journey last year and I am happy to see that I am doing the exact courses + Andrej's series.
Really high quality vid with very useful and good structured information! Keep going and you will be big on yt
I really appreciate it!! Thank you very much 💛
This is brilliant! Thank you so much Boris - definitely needed this guidance to get started in ML/DS
🤩
Honestly, the grind never stops. I reimplemented many papers and published my own ones and still feel somehow i am still a beginner forever
A fellow traumatized researcher 🫡
Haha, I feel the same way! Freaking self doubt in this domain! Are you maintaining your implementations notebook somewhere?
the more knowladge you obtain on a subject gives you more facts about its truth so you have better undetstanding of what you really know and what you dont.... that doeasnt mean tho that you are not in a high level in general
it will never stop... two MSC's om System Design Programming and Comp /Network Security, 25y experience in building systems and design network arch learned like 12-15 different prog langs and yupie Im here watching tutorial from a 22-25 y old about ML and what is the best approach to ML in 2024...
@@kpwlek mind blowing🤯! I think learning is driven by curiosity. When we do a deep dive on a particular topic there's always something which we are not learning. And later that might seems interesting which we start learning. For me this is the cycle
i think reading a paper is a great skill at the end . if u can understand what people did how they did it's great checkmark that u have reached a certain point of advance level of ml
definitely 🤩
For the math part I highly recommend the "Mathematics for Machine Learning" book. It covers all the important stuff without going too much into the details (and gives you the foundation in case you later still want to get into those details). Oh, and it's free.
That book sounds great! Haven't had a look at it, but that you very much for the recommendation! 😊
Who are authors?
@@user-dm7wr5gv4i Marc Peter Deisenroth, A. Aldo Faisal, and Cheng Soon Ong
Linear Algebra And Learning From Data (G. Strang) is a great book as well.
I was like cool it starts with linear algebra, then kept scrolling, what are these words....
thank you so much for this video Boris Meinardus, I was lost in starting the journey of ML but with this video seems like clear roadmap. Every second of this video is informational #gold.
bro please keep it up . we need more informative video like this . Thanks alot brother.
You're voice I find easy to listen to and understand. You have a new subscriber.
Thank you for this! Really well put together
Thanks! I‘m really glad you like it :)
Thank you for such a clear roadmap.
This is something I need. Thank you ML Guru!
Some of the realest and best advice out there
Hey Boris, loved the video! I'm starting my AI graduate studies in January and have also contemplated starting a KZhead Channel explaining the subjects I learn. Just wanted to let you know that you're an inspiration for building a brand for yourself and I'm gonna cheer you on!
Thank you so much for the kind words! I hope my future videos will be as helpful as this one 😊 Also, best of luck with your studies!!
Thanks for the video. I'm a master's student majoring in electronic information. Your video solved my anxiety and made me no longer worry about whether I was progressing too slow and why I couldn't understand the thesis. So just study hard. I have seen and used all the things you mentioned. That's enough.
Frankly this is where i really messed up. I learned python and straight up jumped to learning the ML tech stack. Completely ignoring the math. When i got into NN and deep learning, my lack of math knowledge sent me crashing down coz i kept learning math stuff in bits and pieces only for what i was learning in deep learning at the time. It got chaotic and led to lots of failure in tasks and interviews. Finally an interviewer told me that there is no point going advance if i dont know the basics. I went back to learning math properly and while its been really challenging to study with a pen and paper instead of coding all the time... i know i am now on the right path.
As a comsci major who did some AI as a hobby then formally taking a module on it, I'm burnt out due to the same issue. Its very easy to say just rely on some existing neural network architecture, and ignore the fundamental math principles behind it. The thing is the math (e.g deriatives, linear algebra, statistics) is what drives the algorithms in the first place. By know the math, you know what metrics are appropriate, what other approaches that are more efficient can solve the problem quickly. But its intimidating since most of the data are working in arbitrary higher dimensions that what I learnt before, and its hard to visualize, but less connect the math you prove into code.
Yes, I see... I am sure this is a common problem people have. Maths is scary. I feel like really understanding the basic is important, but expecting to understand everything in the first paper you read is pretty much impossible. But, (and this is the important part) since you have learned the basics, you know enough to research the more advanced maths, and if you give yourself enough time, you will learn everything you need step by step. Even if there is a maths concept you didn't understand in the first paper, once it arises in a further one you read, you will surely be better prepared to understand it. If you continue on failing and learning, you will master the maths, and then see, that it really is not too scary after all :)
Software bros that ignore fundamentals ... hilarious.
can u please tell me from where to learn math I mean resources or tutorial I want to start with machine learning
@@raypamber watch the video again
Awesome video , really a raw to think the feasible path and thinking
Thanks for the explanation !
I'm beginning my conversion MSc in human centred AI and Game development in September, having no computer science degree or strong foundation in maths. I got a C in GCSE. It's a weak point for me but I can see it's crucial in this field. This is really inspiring and it's good to know that I will be okay as long as I work hard. Thank you!
Thanks man you're great help as always!!
I‘m happy my content can actually help!! 😊
Finally a good video on how to actually get started thanks a lot man
This is great advice if you want to learn ML for fun. If you want to get an ML job you will need more than understanding the basic math.
I might be jumping the gun a little, but I feel like in the near future polars is going to be the favorite dataframe tool instead of pandas. I definitely recommend learning it as well, it's a lot more intuitive and faster and a lot of frameworks are adopting it.
Thank you! Very helpful!
Thank you for this!
Keep in mind that this is a guide to "'learn ML"' not "' get a job in ML"'.
💯
@@borismeinarduswhat are extra things to get job
They employ you to bring more profit to the company than your salary.
Is that hard to get job in ML? What i need to know?
@@zaraza_5948ml😂
thank you for your valuable information 🤩 it really helps me
I‘m really happy I could help you ☺️☺️
That is pretty funny. I am currently studying to apply for TU Berlin kolleg. So that is why I followed you because I think saw one of your videos about a day of your life being a TU Berlin student. And ngl I've never seen another video of you. Untill now, about 1 or 2 year(s) later. And this time I was looking for the roadmap for machine learning. And when I was watching your video. I was so confused that where did I see you. And after a little searching I found out why. So the path I am taking is so similar to yours. That makes me not feeling lonely ❤❤
Great video. ML is not about taking a dataset and training random models on it. Every model is different in their own way and the ability to understand the math behind the models helps you determine which model may fit best with your dataset. While these tutorials are great, I’d recommend getting a graduate degree MS/PhD. Most ML positions require a graduate degree, which force you to truly understand the theory
Yes so i am looking for ML for Business Analyst. Can you tell me any resources?
brother, hope you gain more. thanks
Your contents are very appreciated.
Awesome Video!!!🔥 Looking Forward for more such amazing videos ❤🙌
Thank you so much 🤩 I‘ll do my best to create more valuable content ☺️
I wish you had posted this 6 months ago but anyways I can use this video for revision roadmap thanks mate keep it up👏
I was literally thinking about making a proyect of ML but i didnt know where to start and then i saw this video on my recommendations.
Thank you so much for the advice and recommendations :D
You're so welcome!
Awesome video, I just subscribed!
Starting from my point, I think dont rely on GPT too heavily, though it really helps when you're stuck in some problems. Sometimes you have to create your own constrcution of neural networks based on all the basics from python, torch, numpy etc. And your own construction is way more straightforward than the GPT gives. (My experience learning DL for 2 months😜)
Thanks a lot for this. I am about to begin my ML journey and this video showed me the way.
I‘m really happy it could give you a little bit of guidance 🤗
Can we start together... By sharing knowledge with each other and helping it could be helpful for both what do you think
@@shani8175 sure
Which year of college you are in?
@@ananyagupta321 last semester :) have had some non-serious forage into ML earlier but doing it now seriously.
Junior dev bait alert
I was scared that I didn't know enough math after taking linear algebra, but that covers most. I totally agree you can always go back and learn the holes in your knowledge in stead of spending months filling in potential holes.
Good stuff thanks mate
Thanks for this no none sense guide for a complete beginner like myself ❤
Nice one brother... being in the field i couldnt reccomend anything more. Arch are getting quite irrelavant now with these foundation models but have to get the base straight to start debugging something as base math remains same.
thank you, this video is really helpful.🎉
🤗 I‘m really glad it helped you!!
good video, you are a great inspiration for me, congratulations, I am already your new subscriber 👍 blessings 🙏🏻
It's a helpful video for those who don't know how to get started with ML.
I am not sure that basic math such as linear algebra, probabilty in the level of high school, first year of university is enough. I guess some day you will encounter something like of a glass ceiling because advanced ml requires far more advanced math. I am not discouraging the reader to lower his arms, but if you have a choice, it is better to join university where math is taught on above average level. I have a really strong background in pure math, but even with it i sometimes strugge. For e.g. you might need subjects which are really hard to understand such as functional analysis, stochastic processes, descrete math and etc(subjects which is not present in standard engineering math curriculum). But i am working in the research, not on the ml engineer position, so i guess i'm a little bit biased :) But, once again, my advise is to study math as much and as thoroughly as possible. Hope it helps
Very valuable video. Thanks!
thanks, this is very helpful
Hi Boris, thank you so much for sharing your knowledge. I really want to implement a paper like you suggest, not sure where to start? also what kind of computer would you suggest for an ML engineer or DS ? Thanks!
This is brilliant 👏 indeed useful
I know Python and I've taken math up to linear alg (aced all finals in math too) I just need to brush up on my Python and I should be able to deep dive into that ML developer stack!
It's an ok list but it lacks one key element: MLOps. If you don't know how to deploy your model nobody will ever use it - as simple as that. Granted - in a very ML mature company (and those are very, very rare) you will have ML engineers who will take care of deployment but you still will need to know how the MLOps lifecycle works and what part you play as ML developer.
I am starting my journey of ML today and after 3 months I will update how's it going .I will follow the path and I have faith in myself(also in your words ☺) I will get better and will definitely become the best. Also wanna say thank you for your videos it helps a lot.
Way to go! See you in 3 months 🫡🚀
@@borismeinardus Hey there! good to hear about your journey. Even I am getting started on ML. I was wondering if we could collab and learn together by sharing knowledge and help out each other in downs.
@@Unbroken.005 I am down on this
How do I contact you guys
@@olamijutomiwa8336 no need bro in these 3 months I am gonna make python and math strong and learn base of ml
Super useful video. Thank you!
I‘m really happy you liked it! 🤗
I am a Cloud Engineer and was thinking of switching to ML/AI. This video was kinda what I was looking for to get an overview of everything and the roadmap and resources to pick! Thanks man! Keep it up. PS: Liking and Subscribing!
ML is too huge, find out the opportunities where your expertise intersects with ML. Maybe MLOps, or something else!
Same! Planning to work on a MLOps by the end of the year. I find it challenging on how much ML I need for MLOps without becoming ML Engineer
Same situation here, currently in Cloud Devops. Now thinking of moving towards AIOps, MlOps with Python programming for future job security and progress.
Sounds like a solid plan for my 2024, ty
How's it coming so far?
Thanks for this post
thank you , i want to take your opinion about : which one should i learn first ML or Ai ?
First I learned Linear Algebra on yt with Strang course. It took me half a year. Now I'm on the challenge to comple all Khan Academy HS and college courses. It's already half a year in. I still have lots to do. Then I'll take specialized math for ML course. You'd say it's overkill and I waste time. You probably would be right. Also I forget the most of what I learn. But I just want to tick off all the boxes, so whenever math difficult comes in, I wouldn't be like "oh I have no idea what's going on"
I think you're just doing unnecessarily too much. Just learn enough to get you started. Don't let imposter syndrome win over you. The thing about it is that if you understood some basics you will still revisit the maths and stats as you're working on projects
Very helpful and inspirational video, love from TU!
Thank you!! ☺️💛
Andrew Ng's Machine Learning Specialization course is only free for 7 days. After that you have to pay $50 per month to continue.
Thank you so much. This was a great video!
Glad you enjoyed it!
Thank you very much Boris! What sort of side projects did you build to showcase what you know?
I have a video that might answer your question coming out next week on Sunday 😉🤗
Hi, I have learned TensorFlow Keras API to develop ANN & CNN models. I have Question? What is the key difference between TensorFlow & Pytorch. Does Pytorch has more efficient or TensorFlow has more usage cases? like that. Thank you.
Hi Boris, great video! When you re-implement papers do you only do this for papers that provide their code on GitHub?
Thank you for the video. The Machine Learning Specialization course is now 49 USD/month.
Awesome! Thanks
Really helpful keep it going ❤
thanks!!! 😊
Hey! Thank you for this guide. However, Andrew Ng's specialization's free audit version doesn't include notebooks and assignments and things, just the video content. How do you suggest I get some hands-on practice through other means? I don't know yet if I want to invest so much money into the courses.
Thanks for the informative video👍🙂
Glad it was helpful!
Hey Boris, very informative video, I am currently in my 1st year of Comp Sci engineering degree and I am taking Cs50Ai can you suggest me what course should I take next or what is it that I should definitely be doing ?
Borya, I subscribed
Thanks for the overview and orientation. AI learning is the next thing I want to go into.
Go for it! It is very fun and rewarding. At least in my opinion :)
Excellent tutorial, starting that course as we speak
Andre Ks NN series is a goldmine.
💯
Thanks a lot for this
My pleasure 😇
Stark Brudi, weiter so!
🚀🚀🚀
pure gold, thank you! what kind of roles can I apply after those courses?
Thanks 🤗 You can in theory apply to every ML opening you see, but at some level of company the courses alone will not be enough. You will need to demonstrate skills through projects and experience, that‘s why I added the last point in the video :) But there are of course other ways to stand out and I‘ll actually upload a new video covering more tips on exactly that on Sunday. Perhaps that video might be useful to you ☺️
2:14 I don't remember by calculus class looking like that! :)
Great video. Thanks for sharing 🎉
Thanks for watching! 🥳
Thanks Boris, it was indeed very helpful. I have a post graduate degree in physics, but after the degree I really had a terrible life, went through a break up, my business went down due to COVID and i fell into a web of addictions, currently i am trying to overcome my addictions as well as i am under treatment for my depression and anger management, that's why i got myself admitted for a post graduate degree for computer science. I really don't know what i am going to do next. But can you help people like me with math or physics background?
Love it 😊
Yay 🤩 Thank you very much for your amazing content as well! Your videos definitely help a lot in understanding complex ML/ DL concepts!
thanks griezmann!
Thank you sir
Would you do a video with a full practical project using all these tools?
Just love the way you say 'Mauths'!:)
🫣🫣🫣
Wow an amazing video!! Thank you ... When you say Reimplementing a paper and recreating the results what's that supposed to mean? What kinds do you suggest? What are the actual ones that have the chances to have an impact or basically to make a difference (basically solving a problem maybe) And finally what do you recommend?😊😊
I'd imagine he means academic papers where machine learning is used to get some result. For learning purposes, very old ML papers introducing what are now considered basic and standard ML algorithms would perhaps be good, but also modern papers from non ML subjects where older, basic ML is being used to solve a problem in those subjects. The latter are also more likely to have accessible data sets. Modern ML papers are likely to be about cutting edge research into improving existing algorithms or inventing new ones and probably less useful as a beginner. To reimplement a paper, read the paper a few times to understand what they are doing, get hold of the same or similar data, and see if you can implement the machine learning techniques used in the paper to get the same or similar results.
@@johnlally3506 ok thank you very much