In this tutorial, the presenter (Codingdidi) explains why a solid grasp of mathematics is crucial for building, debugging, and improving machine‑learning, deep‑learning, and large‑language‑model systems. Rather than memorising formulas, the focus is on developing intuition about how mathematical concepts underpin algorithms such as linear regression, gradient‑descent optimisation, and transformer architectures. The video outlines a concrete learning roadmap, starting with linear algebra (vectors, matrices, dot products, matrix multiplication, transpose, inverse, determinant, rank) as the foundation for data representation and tensor operations. It then moves to calculus and multivariable calculus (derivatives, chain rule, gradient, Jacobian, Hessian, optimisation, back‑propagation) for understanding loss minimisation. Probability and statistics are presented as essential for handling uncertainty, randomisation, hypothesis testing, and model evaluation. Advanced topics such as eigenvalues/eigenvectors, numerical linear algebra (LU decomposition, SVD), tensors, and discrete mathematics/information theory are highlighted for deeper model analysis and research. The presenter stresses learning through visualisation, manual problem‑solving, coding implementations with NumPy/Pandas, and teaching the material to others. A GitHub repository containing a markdown guide and three free online resources (MIT OpenCourseWare, overfitting.io, an interactive matrix‑visualisation site) are provided for self‑study. The overall message is an encouraging, intuition‑driven pathway to acquire the mathematics needed for AI without heavy coding or formula memorisation.
Chapters
Okay, welcome to another part of this video series, Machine Learning with Python and with me. We are going to discuss mathematics for machine learning. There are so many different concepts of mathematics which are required for machine learning, which are required for machine learning, deep learning, LLM models, or transformer architecture to work, right? So we are going to learn about mathematic concept. We are not going to purely understand how, you know, the formulas are and all those things, but we will be understanding what is required, what are the topics that you have to cover, from where you can learn it from. Because not everyone can be expert in everything, right? So I believe I cannot explain you mathematics in the better way, but I can explain you Python and machine learning in the better way. So therefore, I'm going to give you the recommendation for mathematics and what all are required. Now, first understand why— whenever we give any data to machine learning model, what comes between— in between is the maths. Without maths, machine learning models cannot work well, cannot get onto, you know, understanding the pattern. To make the model understand the pattern, you have to add maths function, then only algorithm work and the prediction will happen. There are multiple different reasons why you should focus on mathematics. First thing, it will help you understand how exactly any algorithm really works. Yes, that's true, because any algorithm— let's say I'm talking about linear regression— how does linear regression work? What is the equation? How does that straight line has to be created? So all of those algorithm behind the scene, if you wanted to understand, then you have to understand mathematics before getting onto the model. Second, you have to also understand mathematics Because we in machine learning do not have to memorize, because we are making model memorize, right? Why should we memorize, right? So that is one reason you have to work on intuition. You do not have to understand, you do not have to basically learn the formulas, no, because there is not going to be any exam. You have to understand the intuition that how exactly this maths formula is applied within machine learning and how does these two points connect. And that's it, that's the second point that to understand. Third, when which, when exactly, like let's say any problem statement comes up in the hackathon, let's say I have to create a machine learning model out of it. So which model do you have to choose? Which one is the right model? Or when it comes to the point that, okay, the model is not behaving well, the model is not giving good accuracy, good precision, then what to do at that point of time? Math is gonna help you decide the right model and right way to train it or tune it. So that is also one reason. Second— fourth, fourth is debugging. When you wanted to, you know, build a model, obviously this is not going to be a smooth journey. So when you want— when you're stuck in this journey, what to do at that point of time? Math is gonna help you to decide. And obviously for research, deep learning model, and advanced topics, it is very, very important. When we will discuss about large language models, how architecture of transformer model works, right? Math is something— how I solve, why this function. So just why behind each and everything will help you. You do not have to memorize anything. You do not have to solve 1,000 questions from one, one topic like calculus, linear algebra. No, you do not have to do that. You just have to understand the intuition behind how does that function works. If I talk about horizontal video and what vertical video, so these videos are nothing but a matrix transformation. So you just have to understand that perspective, that okay, why and how mathematics are used in machine learning and what type of concepts are required for the machine learning models. And that's it. After we got to know that okay, this is why we have to learn, now what to learn? The very first thing that you have to learn is linear algebra. In linear algebra, what you have to study? You have to study matrix, vectors. You have to study vectors, matrix, linear transformations. And majorly, if I talk about main topics, then go for vector matrix, dot product, matrix multiplication, transpose, inverse, determinant, and ranking. Why is this important? Very first, machine learning algorithm has the foundation on these vectrix form. Now the next thing is why one should learn mathematics. We have— now the next thing is we have to learn mathematics, we know that, but what to learn? So the very first concept that you have to learn is the linear algebra. Within linear— now next is what to learn in math. So when it comes to what to learn, the very first topic that you have to cover is the linear algebra. In linear algebra specifically, what do you have to study? You have to study 3 things. Vector, matrix, and linear transform. Now, what are the main topics that you have to cover? It is vector, matrix, dot product, matrix multiplication, transpose, inverse of the matrix, determinant, and rank in matrix. Why to learn this? Because all the machine learning models have the foundation on the same things. As I have given you the example of vertical and horizontal screening or video recording or something, right? This happens with a matrix transformation only. Now, the data generally gets stored in matrix and tensors only in machine learning. The data type which are generally used in any LLM models Any transformer models, any ChatGPT kind of models, the data gets stored in metrics and tensors. Now it is used in some techniques like PCA, it is used in regression model, neural network training, and all those things. And even when there are so many dimensions, let's say I'm working with 5D data, what is the point of working with 5D data, right? When I want to reduce it, I need maths. Without maths, I cannot do it. Next is calculus. What is calculus? In calculus, we have to understand the, uh, derivative, slope, integral part, right? So you basically have to understand the chain rule, gradient, integral. So this is basically for gradient descent type of models, you have to understand backpropagation. How does that work, right? When you wanted to minimize the loss, this will help you. When you wanted, uh, even you want your model to learn from the data, only these derivative or integral part will help your data to summarize, to expand, and all those things with this, right? Multivariable calculus, you have to understand, have to basically understand the functions which are in calculus, like partial derivative, gradient, Jacobian, Hessian, optimization. So basically, whenever we train any complex model, let's say you wanted to create any nano GPT, mini GPT, right? I wanted to create any mini LLM, right, or mini version of any model. So you have to understand minima, maxima, optimizations, and all those things. Obviously we need probability. Probability is used in business analysis, in the role of business analyst, data analyst, machine learning engineers, data science, everywhere. So probability is again important because, uh, the requirement of, you know, randomizing the data when you wanted to randomize it, randomness within the data, uncertainty., right? All those things are required in machine learning models. Then comes the next, which is statistics. Obviously, for collecting the data, for analyzing the data, and for interpreting the data in the better way, we need statistics— mean, variance, standard deviation, distribution, hypothesis testing, correlations. Everywhere statistics is required. Statistics is something that makes your model learn what are the patterns. When you wanted to understand the data patterns, A/B testing, feature selections, evaluation of the model— everywhere it is required. Then comes linear regression. We will cover the linear regression in these videos itself, but for better understanding, I would recommend you to Explore the other websites as well. Underfitting, overfitting kind of stages. Then vector calculus, optimization, eigenvalues and eigenvectors. This is very, very important. Some data structures are important. That is why eigenvalue and eigenvectors are important. You also have to understand the tensors in the first part. Then comes a new numerical algebra where you have to try to solve some linear equations like LU decomposition, SVD, or matrix vectorization. If you will search on YouTube or Aditya other Google, you know, sites and all, you will get it. Although I will be sharing a resource in the end of this video, by the end, and you will get everything. Now the next is the discrete math and the information theory. You basically have to quantify the information, as I every time recommend it. Whenever you are adding the information in your resume, you have to quantify it, right? So same way for model also, number works well, so it is important. Now this is a reminder just that math is not about formula. You do not have to worry about formula. Do not try to learn the formula. It is just about building the intuition. You have to work on building intuition that how these two points relate— data and machine learning, maths and machine learning. How does these two points relate? That behind the scene, how machine learning works. So this is just a reminder which is required for everyone to understand. Okay, the pro tip is whenever you go on to any video, just try to understand it visually, right? Second, when you are understanding it, try to solve it once manually with hand. Do not have to memorize, write down the same thing dictated in the video or, or in the tutorial or the Google or something somewhere, right? So just try to solve it once by your hand. Once you will do it once, you'll remember it forever. Now next is you have to implement the same thing in math. So let's say you are learning about math, saying you have to implement it in code. Basically, now let's say you are learning about matrix, tensors, and vector. Now NumPy is a library where you can create all three of them. NumPy, you can create all three of them. So you have to implement it in code. Although in this machine learning series, I will be telling you how to implement. We will be understanding NumPy, Pandas, all these libraries from scratch itself, but For that, you have to understand what is matrix, what is vector, what is tensor, and how linear equations are solved, and all those basic stuff of mathematics. I hope you already know from your university or from your schooling, but you have to, right? Next is you have to repeat, try to teach it to someone else so that you can remember it for your entire life. You do not have to go revise it. So this is my hack of learning. Whenever I learn something, I try to teach it to someone, and whenever I teach to someone that get stored in my permanent memory and I do not have to, you know, revise it. What I have done is I have created an MD file in my GitHub repo. On, on this GitHub repo, if you will go, you will find all the details about mathematics, its concept, what do you have to exactly learn, and from where you can learn it for free. I have added the links also, so do check it out. The link is already in the description. The link is already in the description. So here is a bonus part of the maths video. I'm going to share 3 resources from which you can learn mathematics effectively. The one is OpenCourseWare, particularly for mathematics. You can get onto the mini lectures and can understand all of the maths topics which are required for machine learning. Next is zero-shot text classification, basically overfitting.io. On this website, you can go and understand any of the maths concepts visually, that how exactly gradient descent works, how it looks like behind the scene, how it works, everything. So this is overfitting.io, the second one. The third one is this one. This is, I think, the new website. And yeah, if you go to the Start Exploring, you will get these topics. And you can, let's say, I want to understand this matrix multiplication. So you can easily, you know, change the parameters from here and can visually see how the matrix multiplication happens. And you, as I told you, right, you have to just get onto the intuition. You do not have to worry about how formulas works, how, what, you know, manually how to do it. No, we're going to do it all with code, but you should have the intuition. So these 3 websites will help you with the interview. So that's it for this video. I hope you have learned and you have got your resources which are required. So happy learning, keep growing, and keep doing.
