Machine Learning Engineer Full Course 2026 | Machine Learning Tutorial For Beginners | Simplilearn

Hi there, welcome to complete machine learning engineer full course. In today's AIdriven world, machines are no longer just following instructions. They are learning from data and improving over time. From recommendation systems to fraud detections to self-driving cars and intelligent chatbots, machine learning powers many of the technologies we use every day. Machine learning engineering goes beyond theory. It involves working with data, building models, optimization, performance, and also deploying solutions that can handle real world scale and complexity. In this complete machine learning engineer course, you will learn how to move from foundational concepts to build and deploying machine learning systems. We'll start with the basics of Python, mathematics, statistics, and then progress into core machine learning algorithms, model evaluation, and optimization techniques. As you advance, you will also explore important areas such as feature engineering, deep learning, model deployment, and envelopes practices. You'll also gain hands-on experience with industry tools, and libraries used to build and manage machine learning workflows. By the end of this course, you'll have a strong understanding of how machine learning models are built, evaluated, and deployed in production environments. Having said that, let's take a look at today's agenda for this course. We'll start off with module one which is introduction to machine learning and AI. Module two is Python programming for machine learning. Module three is mathematics and statistics for ML. Module four is data collection and data prep-processing. Module five is exploratory data analysis. Module six is feature engineering and feature selection. Module seven is supervised learning algorithms. Module 8 is unsupervised learning algorithms. Module 9 is model evaluation and performance metrics. Module 10 is introduction to deep learning. Module 11 is model deployment basics. Module 12 is envelopes and model life cycle management. Module 13 is real world machine learning projects. Module 14 is interview question and career guidance. Hope I made myself clear with this agenda. That said, if these are the type of videos you'd like to watch, then hit that subscribe button with the bell icon to get notified whenever we post. Also, just so that you know, if you want to upskill yourself, master generative AI and land your dream job or even grow in your career, then you must explore SimplyLearn's cohort of various generative AI training and professional certificate programs. Simply offers a variety of master certification and post-graduate programs in collaboration with some of the world's leading universities. Through our courses, you will gain knowledge along with work ready expertise in skills like Python, Agentic AI, AI automation systems, LLMS, and over a dozen others. And that's not all. You'll also get the opportunity to work on multiple projects led by industry experts working in top tier service- based and product companies. After completing these courses, thousands of learners have transitioned into AI and machine learning role as a fresher or moved on to a higher paying job and profile. If you're passionate about making your career in this field, then make sure to check out the link in the pin comments and in the description box to find an AI and machine learning program that fits your experience and areas of interest. So let's get started with our machine learning engineer full course with a small quiz. Which type of learning uses label data? Is it unsupervised learning, supervised learning, reinforcement learning, deep learning? Please let us know your answers in the comment section below. Now over to our training experts. We know humans learn from their past experiences and machines follow instructions given by humans. But what if humans can train the machines to learn from their past data and do what humans can do and much faster? Well, that's called machine learning. But it's a lot more than just learning. It's also about understanding and reasoning. So today we will learn about the basics of machine learning. So that's Paul. He loves listening to new songs. He either likes them or dislikes them. Paul decides this on the basis of the song's tempo, genre, intensity, and the gender of voice. For simplicity, let's just use tempo and intensity for now. So, here tempo is on the x-axis, ranging from relaxed to fast, whereas intensity is on the y-axis, ranging from light to soaring. We see that Paul likes the song with fast tempo and soaring intensity while he dislikes the song with relaxed tempo and light intensity. So now we know Paul's choices. Let's say Paul listens to a new song. Let's name it as song A. Song A has fast tempo and a soaring intensity. So it lies somewhere here. Looking at the data, can you guess whether Paul will like the song or not? Correct. So Paul likes this song. By looking at Paul's past choices, we were able to classify the unknown song very easily, right? Let's say now Paul listens to a new song. Let's label it as song B. So song B lies somewhere here with medium tempo and medium intensity. Neither relaxed nor fast, neither light nor soaring. Now, can you guess whether Paul likes it or not? Not able to guess whether Paul will like it or dislike it. Are the choices unclear? Correct. We could easily classify song A. But when the choice became complicated as in the case of song B. Yes. And that's where machine learning comes in. Let's see how. In the same example for song B, if we draw a circle around the song B, we see that there are four votes for like whereas one vote for dislike. If we go for the majority votes, we can say that Paul will definitely like the song. That's all. This was a basic machine learning algorithm also. It's called K nearest neighbors. So this is just a small example in one of the many machine learning algorithms quite easy right believe me it is but what happens when the choices become complicated as in the case of song B that's when machine learning comes in it learns the data builds the prediction model and when the new data point comes in it can easily predict for it more the data better the model higher will be the accuracy there are many ways in which the machine learns it could be either supervised learning unsupervised learning or reinforcement learning. Let's first quickly understand supervised learning. Suppose your friend gives you 1 million coins of three different currencies. Say 1 rupee, 1 and 1 dirham. Each coin has different weights. For example, a coin of 1 rupee weighs 3 g. 1 euro weighs 7 g and 1 dirham weighs 4 g. Your model will predict the currency of the coin. Here your weight becomes the feature of coins while currency becomes their label. When you feed this data to the machine learning model, it learns which feature is associated with which label. For example, it will learn that if a coin is of 3 g, it will be a 1 rupee coin. Let's give a new coin to the machine. On the basis of the weight of the new coin, your model will predict the currency. Hence, supervised learning uses labeled data to train the model. Here, the machine knew the features of the object and also the labels associated with those features. On this note, let's move to unsupervised learning and see the difference. Suppose you have cricket data set of various players with their respective scores and the wickets taken. When we feed this data set to the machine, the machine identifies the pattern of player performance. So, it plots this data with the respective wickets on the x-axis while runs on the y-axis. While looking at the data, you'll clearly see that there are two clusters. The one cluster are the players who scored high runs and took less wickets while the other cluster is of the players who scored less runs but took many wickets. So here we interpret these two clusters as batsmen and bowlers. The important point to note here is that there were no labels of batsmen and bowlers. Hence the learning with unlabeled data is unsupervised learning. So we saw supervised learning where the data was labeled and the unsupervised learning where the data was unlabeled. And then there is reinforcement learning which is a reward-based learning or we can say that it works on the principle of feedback. Here let's say you provide the system with an image of a dog and ask it to identify it. The system identifies it as a cat. So you give a negative feedback to the machine saying that it's a dog's image. The machine will learn from the feedback and finally if it comes across any other image of a dog, it'll be able to classify it correctly. That is reinforcement learning. To generalize machine learning model, let's see a flowchart. Input is given to a machine learning model which then gives the output according to the algorithm applied. If it's right, we take the output as our final result. Else we provide feedback to the training model and ask it to predict until it learns. I hope you've understood supervised and unsupervised learning. So let's have a quick quiz. You have to determine whether the given scenarios uses supervised or unsupervised learning. Simple, right? Scenario one. Facebook recognizes your friend in a picture from an album of tagged photographs. Scenario two, Netflix recommends new movies based on someone's past movie choices. Scenario three, analyzing bank data for suspicious transactions and flagging the fraud transactions. Think wisely and comment below your answers. Moving on, don't you sometimes wonder how is machine learning possible in today's era? Well, that's because today we have humongous data available. Everybody's online either making a transaction or just surfing the internet and that's generating a huge amount of data every minute and that data my friend is the key to analysis. Also, the memory handling capabilities of computers have largely increased which helps them to process such huge amount of data at hand without any delay. And yes, computers now have great computational powers. So there are a lot of applications of machine learning out there. To name a few, machine learning is used in healthcare where diagnostics are predicted for doctor's review. The sentiment analysis that the tech giants are doing on social media is another interesting application of machine learning. Fraud detection in the finance sector and also to predict customer churn in the e-commerce sector. While booking a cab, you must have encountered search pricing often where it says the fair of your trip has been updated. Continue booking. Yes, please. I'm getting late for office. Well, that's an interesting machine learning model which is used by global taxi giant Uber and others where they have differential pricing in real time based on demand, the number of cars available, bad weather, rush hour, etc. So they use the search pricing model to ensure that those who need a cab can get one. Also, it uses predictive modeling to predict where the demand will be high with a goal that drivers can take care of the demand and search pricing can be minimized. Great. Hey Siri, can you remind me to book a cab at 6 p.m. today? Okay, I'll remind you. Thanks. No problem. Comment below some interesting everyday examples around you where machines are learning and doing amazing jobs. So, that's all for machine learning basics today from my side. Keep watching this space for more interesting videos. Until then, happy learning. Artificial intelligence, machine learning, and deep learning represent the evolution of computer science towards creating intelligent systems. AI is the broader concept striving to build machines capable of humanlike intelligence. ML is a subset of AI emphasizing algorithms that learn from data to make predictions or decisions. DL in turn is a specialized branch of ML that employs deep neural networks to model complex patterns. Imagine an AI powered voice assistant like Apple Siri. It utilizes ML to understand and respond to user queries, learning from interactions over time. Deep learning comes into play when Siri recognizes speech patterns or interprets natural language using neural networks to process intricate features. The better it becomes at understanding diverse accents or refining responses. Exemplifying the continuous learning inherent in these technologies. AI seeks to emulate human intelligence. ML harness data for learning and DL employs deep neural networks for intricate task. The integration of these technologies manifest in everyday applications transforming how we interact with and benefit from intelligent systems. This technology enables voice interaction allowing the device to play music, set alarms, present audio books and provide upto-date information on topics like news, weather, sports and traffic reports etc. Let's move forward and see what is machine learning. Machine learning is a subset of artificial intelligence that focuses on developing algorithms and models capable of learning and making predictions or decisions without being explicitly programmed. ML systems leverage data to recognize patterns, adapt and improve their performance over time. There are several types of machine learning. Number one, supervised learning. The algorithm is trained on a label data set where each input is associated with a corresponding output. Number two comes as unsupervised learning. Unsupervised learning deals with unlabelled data to find inherent patterns or structures within the information. And then comes the reinforcement learning. This type involves training agents to make sequences of decisions by interacting with an environment. And then comes semi-supervised learning. Semi-supervised learning combines supervised and unsupervised learning elements typically using a small amount of labelled data and a larger pool of unlabelled data. Let us move forward and see what deep learning is. Deep learning, a branch of machine learning, focuses on algorithms inspired by the human brain structure and functionality. It excels in processing vast amounts of both structured and unstructured data. At the heart of deep learning are artificial neural networks, empowering machines to make decisions. The key distinction between deep learning and machine learning lies in data presentation. Machine learning algorithms typically demand structured data while deep learning networks operate through multiple layers of artificial neural networks allowing them to handle diverse data formats. So let's start with the difference between artificial intelligence, machine learning and deep learning. And this we'll show in a table form. So starting with the definition. So definition of artificial intelligence. So broad field of machine learning or creating machines with intelligent behavior is artificial intelligence. And when we talk about machine learning, it's the subset of AI focusing on algorithms learning from data. And then comes the deep learning that is specialized subset of ML using deep neural networks. And now we'll see the difference with the learning approach between all these three. So in learning approach artificial intelligence can include rule-based systems, expert system and more. And in machine learning it learns from data patterns without explicit programming. And then comes the deep learning where it learns hierarchical representation using neural networks. And if we talk about scope, it encompasses various techniques beyond learning from data. And in machine learning, it primarily focus on learning patterns from data. And then the deep learning, it specifically utilizes deep neural networks for complex task. And now we'll move to the next difference and we'll start with an example. So in artificial intelligence, the example is autonomous vehicles, chatboards or expert systems. And for machine learning, it's spam filters, recommendation systems, image recognition. And in deep learning, it is image and speech recognition, natural language processing. And now we'll see the difference for the data requirements. So it depends on the specific application and problem solving approach. And in machine learning, it requires labeled or unlabelled data or training. And for the deep learning, it relies on large amounts of labelled data for training deep networks. And now for the complexity, artificial intelligence addresses a wide range of task including those beyond ML. And in machine learning, it deals with moderate to complex task depending on algorithms. And for the deep learning, it is well suited for intricate task often requiring substantial computational resources. And now see the flexibility. So for the artificial intelligence, it can be rule-based, evolving and adaptive. And for the machine learning, the flexibility adapts to patterns and the changes in data. And for the deep learning, it adapts to hierarchical representations and diverse data types. And then comes the training process. So in artificial intelligence, training process varies based on specific AI techniques used. And in machine learning, training involves feeding data and adjusting model parameters. And in deep learning, training involves optimizing neural weights and structures. And now we'll talk about the applications between all these three terms that is A IML and deep learning. So for artificial intelligence the applications are robotics, natural language processing, game playing. And for machine learning it's predictive analytics, fraud detection and healthcare diagnostic and for the deep learning that is image recognition, speech synthesis and language translation. We have Rahul who will take you through the various applications of machine learning and how you can be a machine learning engineer. Machine learning has improved our lives in a number of wonderful ways. Today let's talk about some of these. I'm Rahul from SimplyLearn and these are the top 10 applications of machine learning. First let's talk about virtual personal assistants. Google Assistant, Alexa, Cortana and Siri. Now we've all used one of these at least at some point in our lives. Now these help improve our lives in a great number of ways. For example, you could tell them to call someone. You could tell them to play some music. You could tell them to even schedule an appointment. So, how do these things actually work? First, they record whatever you're saying, send it over to a server, which is usually in a cloud, decode it with the help of machine learning and neural networks, and then provide you with an output. So, if you ever notice that these systems don't work very well without the internet, that's because the server couldn't be contacted. Next, let's talk about traffic predictions. Now, say I wanted to travel from Buckingham Palace to Lots Cricket Ground. The first thing I would probably do is to get on Google Maps. So, search it and let's put it here. So, here we have the path you should take to get to Lord Cricket Ground. Now here the map is a combination of red, yellow and blue. Now the blue regions signify a clear road that is you won't encounter traffic there. Yellow indicate that they're slightly congested and red means they're heavily congested. So let's look at the map a different version of the same map. And here as I told you before red means heavily congested, yellow means slowmoving and blue means clear. So how exactly is Google able to tell you that the traffic is clear, slowmoving or heavily congested? So this is with the help of machine learning and with the help of two important measures. First is the average time that's taken on specific days at specific times on that route. The second one is the real-time location data of vehicles from Google Maps and with the help of sensors. Some of the other popular map services are Bing Maps, Maps.me. And here we go. Next up, we have social media personalization. So say I want to buy a drone and I'm on Amazon and I want to buy a DJI Mavic Pro. The thing is it's close to one lap, so I don't want to buy it right now. But the next time I'm on Facebook, I'll see an advertisement for the product. Next time I'm on YouTube, I'll see an advertisement. Even on Instagram, I'll see an advertisement. So here with the help of machine learning, Google has understood that I'm interested in this particular product. Hence, it's targeting me with these advertisements. This is also with the help of machine learning. Let's talk about email spam filtering. Now, this is a spam that's in my inbox. Now, how does Gmail know what's spam and what's not spam? So, Gmail has an entire collection of emails which have already been labeled as spam or not spam. So after analyzing this data, Gmail is able to find some characteristics like the word lottery or winner. From then on, any new email that comes to your inbox goes through a few spam filters to decide whether it's spam or not. Now, some of the popular spam filters that Gmail uses is content filters, header filters, general blacklist filters, and so on. Next, we have online fraud detection. Now, there are several ways that online fraud can take place. For example, there's identity theft where they steal your identity. fake accounts where these accounts only last for how long the transaction takes place and stop existing after that and man-in-the-middle attacks where they steal your money while the transaction is taking place. The feed forward neural network helps determine whether a transaction is genuine or fraudulent. So what happens with feed forward neural networks are that the outputs are converted into hash values and these values become the inputs for the next round. So for every real transaction that takes place there's a specific pattern. A fraudulent transaction would stand out because of the significant changes that it would cause with the hash values. Stock market trading machine learning is used extensively when it comes to stock market trading. Now you have stock market indices like Nikai. They use long short-term memory neural networks. Now these are used to classify, process and predict data when there are time lags of unknown size and duration. Now this is used to predict stock market trends. Assistive medical technology. Now medical technology has been innovated with the help of machine learning. Diagnosing diseases has been easier from which we can create 3D models that can predict where exactly there are lesions in the brain. It works just as well for brain tumors and ice schemic stroke lesions. They can also be used in fetal imaging and cardiac analysis. Now, some of the medical fields that machine learning will help assist in is disease identification, personalized treatment, drug discovery, clinical research, and radiology. And finally, we have automatic translation. Now, say you're in a foreign country and you see billboards and signs that you don't understand. That's where automatic translation comes of help. Now how does automatic translation actually work? The technology behind it is the same as the sequence to sequence learning which is the same thing that's used with chatbots. Here the image recognition happens using convolutional neural networks and the text is identified using optical character recognition. Furthermore, the sequence to sequence algorithm is also used to translate the text from one language to the other. And today I'm going to tell you how you can become a machine learning engineer. But before we begin, let me tell you what a machine learning engineer actually does. A machine learning engineer creates and maintains machine learning solutions to solve business problems. They constantly tweak and optimize the solutions for maximum performance and scalability. They solve business problems like reducing customer churn, running targeted marketing campaigns, and improving product experience. They also help with predicting whether a particular hypothesis will be profitable in the future. They contribute to cutting edge research in AI and machine learning. Now before you can start off on your journey to becoming a machine learning engineer, there's a certain number of steps that you need to follow. On this learning path, your first step is to improve your math skills. Mathematics plays a very important role in helping you understand how machine learning and its algorithms work. Among the many concepts that you need to understand, three of the most important ones are probability and statistics, linear algebra, and calculus. Now let's have a detailed look at probability and statistics. Firstly, you have the base theorem. This is used in the nave base algorithm to help categorize your data. Then we have probability distribution. This helps you determine how frequently a particular event will take place. For example, you can determine what your premium for your insurance will be based on the probability distribution of expenditure pertaining to insurance claims. You must also learn how sampling and hypothesis testing works. Now let's look at linear algebra. Now linear algebra has two main concepts, matrices and vectors. They're both used widely in machine learning. Now, matrices are used for image recognition where the entire image that you're using is already in the form of a matrix. You need to be able to work with matrices, perform simple operations like addition, subtraction, multiplication, inverse, transpose and so on. Now, the recommended systems that you see in applications like Netflix or Amazon actually work on vectors. This vector is the customer behavior vectors and they use distance measures. Now, let's have a closer look at calculus. You have differential calculus and integral calculus. Now these help in determining the probability of events. For example, in finding the posterior probability in a navebased model. In your next step to becoming a machine learning engineer, you need to develop good programming skills. And let me tell you, there are a huge number of options from which you can choose. There's Python, there's C, there's C, there's Java and so much more. Now, here's a graph of the job postings from 2014 to 2017. And you can see that there are two languages that dominate since 2015. Python and R. Now these are one of the most wanted languages when it comes to machine learning engineers. These are closely followed by JavaScript and C. Hence, we would like to recommend that you learn Python and R as they're the best options when it comes to coding in machine learning algorithms. So here are a few things that you need to know about Python and R. Python is an object-oriented language which means their main emphasis is on the object. R is a functional language which means their emphasis is on creating and manipulating functions. Python relies on its many packages and R is slightly faster than Python because it has inbuilt packages. Now, Python is generic and is suitable if you need to integrate it with any other software. R works a little more closer with statistical analysis. Your next step is to get yourself some data engineering skills. Now, these skills are important as they help you analyze and process your data as soon as you get it. Now there are three major steps when it comes to data engineering. Firstly, there's data prep-processing. Now these refer to all the steps that you need to perform before data can be processed by the machine learning algorithm. You have cleaning, parsing, correcting and consolidating the data. Then you have ETL or extract, transform and load. You need to know how data can be extracted from the internet or a local server. You need to know how to transform the data. For example, not all formats of data would be accepted by the program. So you need to convert the data into a format that is accepted by the program. Then you need to know how the data is loaded into your program. The final step is to have knowledge about database management softwares or DBMS. You need to be well versed with MySQL, the Oracle database and NoSQL. Your next step is to learn machine learning algorithms. Now among all the machine learning algorithms that you can see on screen, you can divide them into two different categories. ones that fall under supervised machine learning and the others that fall under unsupervised machine learning. Even after such a division, you can further subdivide them into classification and regression algorithms. Now, all algorithms except linear regression fall under the category of classification which is used to determine whether a particular data falls into a particular category. On the other hand, you have linear regression that falls under regression algorithms. This is used to predict a particular value. Then you have K means clustering and hierarchal clustering. Now these fall in the category of clustering. This is used to group data into clusters based on certain similar attributes. Then you have the a priori algorithm that falls under the concept of association. Association is used to determine patterns of association among variables in large data sets. So now that you know about these algorithms, let me tell you where you can learn about them. Let's take a look at our simply learn channel and let's go to playlists. And on this we have a dedicated set of playlists that talk about machine learning. Here you have videos on machine learning. How machine learning is different from deep learning and artificial intelligence. Machine learning with Python, K means clustering, decision trees and so on. If you want to learn more, you could also go through some deep learning algorithms like convolutional neural networks, recurrent neural networks, long short-term memory networks and so on. In fact, we have a detailed playlist that talks about the concepts of deep learning. Here you can find videos on what is deep learning, TensorFlow, what is a neural network, convolutional neural networks, and recurrent neural networks. Now, after you master these algorithms, you need to learn how you can select the right algorithm for your problem. Then you need to create a good model with one or more algorithms. after which you need to keep tweaking and optimizing the model so that you can get the maximum accuracy. For further reading, you can go through GitHub where there's more than 21,000 repositories under machine learning. You can also recreate published research papers adding to your experience of working with machine learning. And now for your last step, learning machine learning frameworks. Machine learning frameworks help make the lives of developers as well as users a whole lot easier. They help remove the complex part of machine learning and make it available for everyone who wants to use it be it developers or other users. Now let's look at the widely used machine learning frameworks. There's TensorFlow, Theiano, Torch, Scikitlearn and so on. Now let's look at some of them in detail. First let's look at TensorFlow. TensorFlow is the most widely used machine learning framework. It's used for machine learning as well as deep learning. Now it's an open-source software library which performs numerical computations which is done with the help of dataf flow graphs. Google translate is one of the most popular use cases of TensorFlow. Now let's look at Theano. Theo helps you define, optimize and evaluate mathematical expressions. It was developed in the University of Montreal. Lasani blocks and keros are its most popular libraries. Now let's look at Spark ML library. Now this is Apache Spark's machine learning component. It also provides libraries for machine learning which are built on top of RDDs or resilient distributed data sets. It's very good at providing iterative computation and provides very high algorithmic performance. Now let's look at scikitlearn. Scikitlearn is able to provide a huge range of supervised as well as unsupervised algorithms for machine learning. It is built on existing libraries like numpy, scypi and mattplot lib. Scikitlearn actually started off as a Google Summer of Code project and now has 23,000 GitHub commits. And that's it. Congratulations, you're now a machine learning engineer. Now, let's look at the job opportunities in machine learning. Now, if you have a look at this graph, it kind of speaks for itself. Before 2015, machine learning was much less popular than big data and cloud computing. But all of that suddenly changed. And right now, a machine learning engineer earns around $114,000 perom. And this is a clear indication that organizations are ready to invest heavily in people who are skilled in the concepts of machine learning. You can also make the learning process easier by using simply learns machine learning certification. Simply learns machine learning certification course provides 36 hours of instructorled training. Provides 25 plus hands-on exercises. gives you practical applications of 15 plus machine learning algorithms and helps you master the concepts of supervised and unsupervised learning. It also introduces you to artificial intelligence, tells you the techniques of machine learning, data processing, regression, classification and so much more. So if you want to take your first step to getting certified and getting ahead offline now machine learning systems are integrated into real world operations and are capable of making decisions that impact the business immediately. Here's an example to make it clearer. In the past, an e-commerce website might use machine learning to predict what products a customer might want based on their past purchases. Now the systems can constantly learn from new customer data continuously refining those predictions in real time as customers preferences are changing. The world of machine learning has evolved from theory to practice and this has created a huge demand for machine learning engineers who can build scalable systems and make them work in real world environments. The impact of machine learning is now directly tied to business outcomes and machine learning engineers are at the center of that transformation. You might be thinking, okay, I get it. Machine learning is impactful, but what exactly does a machine learning engineer do compared to other roles in tech? That's a great question. In the world of machine learning, you'll hear about a few key roles such as data scientist, machine learning, and AI engineer. Let's break them down so you know exactly where you fit in. Data scientists are like the detectives of data. They spend their time analyzing large data sets, finding trends, and trying to extract meaningful insights. They build models but their main focus is usually on data exploration and experimenting with various algorithms. They don't typically focus on deploying those models into production environments. Machine learning engineers on the other hand these are architects. They take the models built by data scientists and build scalable deployable systems. They work on creating solutions that will not only work in the short term but can also scale to handle real world data in massive volumes. The machine learning engineer is responsible for ensuring that machine learning systems are integrated into businesses that can work seamlessly with existing technologies. AI engineers focus more on the application side of things. They build AI powered products like chatbots, voice assistants, and real-time systems. While their work often overlaps with ML engineers, they are typically more focused on the userfacing product and how machine learning fits into it. As an ML engineer, your primary focus is to take models and turn them into actionable solutions that are deployed in real world systems. We shall now move on to why 2026 is the right time to enter machine learning. Now that you know the role of an ML engineer, let's talk about why 2026 is the perfect time for you to jump into the field. You've probably heard that machine learning is a hot topic, but what does that mean for you as someone starting out in this field? So, I'll help you break that down for you. First, the demand for ML engineers has skyrocketed. The world is full of problems that need solving, and machine learning has proven to be one of the most effective tools to solve them. From predicting customer preferences to automating critical business functions, machine learning is changing how businesses operate. Secondly, the tools used to build machine learning systems are more accessible than ever before. In the past, machine learning was viewed as something experimental, something that required a lot of effort just to set up. But today machine learning platforms and frameworks such as TensorFlow, PyTorch and Scikitlearn have matured significantly. These tools make it easier to build and deploy models that can scale to handle real world data. We shall now move on to why is this the right time for you to get started out as a machine learning engineer. So how do you get started? The first step is to build a strong foundation. You might be excited to start building models and diving into algorithms. But before that you need to understand the core concepts that drive all the machine learning systems. These include mathematics, programming and data handling. You don't need to be an expert in all of these areas, but you do need to understand the basics. Think of these as building blocks of everything that you will need to learn machine learning. Let's start with mathematics. You don't need to be a math genius, but you do need to understand the basics. There are three main areas of math that will help you get started out as an ML engineer, and those are linear algebra. This is the study of vectors and matrices which are used to manipulate and process data in machine learning models. If you have heard terms like feature vectors or matrix operations, that's linear algebra at play. This is the core of how machine learning algorithms operate. This is the math behind optimizing models. You'll use calculus to adjust the parameters of machine learning models and minimize the error between predicted and the actual values. Specifically, derivatives are used to find the best way to fit a model into the data. Statistics. Machine learning is all about working with uncertaintity, and statistics will help you make sense of it. Whether you're dealing with probability distributions or hypothesis testing, statistics can help you understand patterns in data and make decisions with uncertain information. These mathematical concepts will help you build a more accurate model and optimize it effectively. Let's move on to programming stack. If you're new to programming, don't worry. Python is the language that you will want to learn. It's simple to get started with and has a huge ecosystem of libraries especially designed for machine learning. The key libraries that you will need to master include numpy. This library is used to handle large arrays of matrices and data which is the backbone of most machine learning algorithms. If you plan on working with large data sets, you'll be using numpy a lot. Pandas. This library is great for data manipulation and analysis. You'll use pandas to clean, organize, and transform data, making it ready for machine learning models. Scikitlearn. This library provides simple, easy to use tools for building machine learning models. It covers everything from data prep-processing models like regression and classification. Once you're comfortable with these tools, you'll be able to start building machine learning models and working with real world data. Next up is SQL, which stands for structured query language. As an ML engineer, you will be working on lots of data, and SQL will help you query databases to retrieve information that you need. You'll use SQL to extract data, filter it, and join tables together, making sure that you have the right data to your models. Once you have your data, the next step is data wrangling. Data wrangling is a huge part of your job, and if you master it, it will save you a lot of time and frustration while building models. We shall now move on to types of machine learning. Let's talk about the different types of machine learning. There are three main categories which are supervised learning, unsupervised learning and reinforcement learning. Speaking of supervised learning, this is when you have label data. You train your model on data where the answers are already known. The goal is for the model to learn the relationship between inputs and outputs so it can predict the future outcomes. Unsupervised learning. In this case, the model works with unlabelled data and it tries to find hidden patterns and groupings in the data. This is useful for tasks like clustering or anomaly detection. Reinforcement learning. This type of learning involves training an agent to make decisions by interacting with its environment and receiving rewards or penalties based on its actions. This is often used in game AI or robotics. We shall now move on to feature engineering and evaluation. After cleaning your data, the next step is feature engineering. This is the process of transforming raw data into meaningful features that help your model make better predictions. Once you've engineered your features, it's time to evaluate the model. You'll use metrics like accuracy, precision, and recall to see how well your model is performing. Proper evaluation ensures that your model is ready for real world applications and that it can handle data in production. We shall now move on to the six-month learning plan in order to become an ML engineer in 2026. So, how do you actually get started? Here's your six-month learning plan to guide your journey. Firstly, focus on learning Python, mathematics, and SQL. Dive into machine learning algorithms and hands-on projects. Then, learn deep learning and work with frameworks like TensorFlow or PyTorch. By the sixth month, you can work on a capstone project that covers the full pipeline from data collection to deployment. We shall now speak about the ML engineer tool stack. So let's talk about the tools that you will need as a machine learning engineer. You must be wondering what tools should I be learning to become a successful ML engineer in 2026. So let's simplify this. Git and GitHub are the first things on your list. At first glance, version control might seem like something that only coders need to worry about. However, it can be really crucial. Why? Because version control allows you to track every change that you make to your code, collaborate with teams, and roll back changes if something goes wrong. Imagine you're working on a huge project and you mess something up. Without version control, you could lose hours of work. But with Git and GitHub, you could go back and fix to any point and time. You might be thinking, okay, I get that Git is useful, but what about the tools that can actually help me build and deploy machine learning models? Now, that's a great question. So let's talk about the cloud platforms like AWS, Google Cloud, and Azure. In the past, machine learning models were often built and tested upon local machines, but we quickly realized that it wasn't scalable. These cloud platforms give you the ability to handle large data sets, run models on highowered servers, and scale them as your projects grow. Instead of relying on your personal laptop to do all of the heavy lifting, you can leverage the cloud to train models much faster and handle massive data sets without worrying about memory limitations. We shall now move on to the next topic which is on MLOps life cycle. You have now got all of your tools in place. But how do you go from all of these tools coming together in the MLOps life cycle? So you must be wondering what does all of the MLOps life cycle tools even look like and how do I manage the process from starting to finish? Well, here's the thing. MLOps is like a welloiled machine. It involves a series of stages that ensure that your models remain reliable, scalable, and adaptable to new data over time. Think of it like a car assembly line. First, you train the car and then you deploy it into the real world. After that, you need to monitor it and see if it runs smoothly. If it breaks down, you will have to retrain it. Let's break it down into key steps. Training your model. This is where you create the model using your training data. This is a very experimental phase where you try different algorithms, tweak parameters, and optimize the model for better performance. Deploying it to production. Now that your model is ready, it's time to put it into action. This means making the model available to users or clients. Whether it's in a mobile app or a web service, deployment is where the magic happens. Monitoring performance. Once your model is live, you can't just forget about it. It's like checking the car's tire pressure after it's been on the road for a while. You need to continuously track how well your model is doing. If it starts to slip or underperform, it's time for tweaks and adjustments. Lastly, we have retraining. Over time, your model may need to be retrained with new data to stay accurate. This is especially true in industries where the environment is constantly changing, like e-commerce or finance. Restraining ensures that the model stays relevant and continues to provide value. You may be wondering that this sounds like a lot of work and that's true. That's where MLOps tools like MLOR help you streamline this process. By automating and managing this life cycle, you can spend less time dealing with the back end and more time focusing on building innovative solutions. We shall now move on to experiment tracking. As you dive deeper into machine learning, you'll quickly realize how important it is to track your experiments. At first, this might seem a little overwhelming. You might be thinking, I can't just run a model and hope for the best, right? But trust me, tracking your experiments is one of the best habits that you can develop early on. Think of it like logging your workout progress. Even if you don't track the results, how will you know whether you're improving or not? The same goes for machine learning. By using experiment tracking tools flow and weights and biases, you can keep a detailed record of every model and every hyperparameter and every evaluation metric. Now imagine you're building a recommendation system for an online store. You try different adjustments, algorithms, and get different results. With experiment tracking, you can easily compare which configurations work best. You can easily compare which configurations work best and learn from the past mistakes. You'll always know which experiment gives you the best results and which needs tweaking. Tracking experiments also helps in collaboration. If you're working with a team, being able to see everyone's experiments in one place makes it easier to understand their approach and build on each other's work. We shall now move on to projects and portfolio. Now that you've got your tools and workflow in place, it's time to focus on building a strong portfolio. You might be thinking, how do I make my portfolio stand out to potential employees? That's a great question. The answer is simple. Real world projects. Think about it. Employers want to see what you can do in practice. They don't just want to see theoretical knowledge. They want to see that you can solve problems and build working systems that can make a difference. So for that reason, we have mini project examples. At this point, you're probably itching to start building something yourself. Well, hands-on projects are the best way to solidify what you've already learned. For example, you could create a customer churn prediction model or a fraud detection system. These are practical real world projects that demonstrate your ability to build solutions from start to finish. Make sure you document your projects clearly on GitHub and always include a detailed explanation of your approach, challenges, and results. Remember, the goal is to show that you understand the problem and build a working solution. We'll now move on to Kaggle and open source. As you continue to build your portfolio, I highly recommend diving into Kaggle competitions and contributing to open source projects. Kaggle is an amazing platform where you can work on real world data sets and solve problems that companies and research institutions are facing. Not only will you improve your skills, but you'll also have the opportunity to see how top data scientists and machine learning engineers approach similar problems. Contributing to open-source projects is another excellent way to showcase your skills. It shows that you can work well with others understanding existing systems and contribute to the community. Plus, it's a great way to gain visibility and make connections with other engineers. We shall now speak about the résumés which work for you. Speaking about your resume, when it comes to landing a job as a machine learning engineer, your resume needs to be focused on real world projects and practical experience. So, be sure to highlight the projects that you've worked on, the tools you've used, and most importantly, the impact that your work has had. If you build a recommendation engine that boosted product sales by 20%, make sure that you include that. Employers want to see how your work contributes to solving real business problems. Don't forget to include links, Kaggle profile, or any other open source contributions that you have made. Employers love seeing code and showing them your projects is the best way to stand out. We shall now speak about what interviewers test in 2026. By now you must be wondering what do employers actually look for in an ML engineer. So in 2026 interviews are not just about technical knowledge. Employers just want to see how well you can communicate your thought process and real world problems. You'll likely face practical tests that challenge you to build a model or analyze data in real time. You'll also be tested on how well you explain your approach and justify the decisions that you made during the project. Prepare to discuss things like why you choose a specific algorithm for a problem or how you handle issues like data imbalance or overfitting along with what metrics you use to evaluate your model's performance. Being able to communicate clearly your process and how you arrived at your solution is a skill that will set you apart from other candidates. We shall now speak up about the ML trends that you will need to follow. Machine learning is evolving fast and staying up to date is key to remaining relevant in this field. So here are a few trends to keep an eye on in 2026. Generative models. These models can generate new data based on patterns they learn and they're used for things like text generation, image creation, and even music composition. AutoML automated machine learning tools are making it easier for non-experts to build machine learning models. As a result, more people will be able to contribute to this field without needing to become experts. Privacy first models. As privacy concerns grow, machine learning models are being designed to work securely and ethically and these don't compromise on user privacy. Staying on top of these trends will help you remain competitive and innovative in the ever evolving field of machine learning. So in conclusion, I will say consistency is key in machine learning. You don't need to know everything right away, but you must stay committed to learning and building. Start small, keep experimenting, and keep improving. The road to become a machine learning engineer may seem long, but with the right tools and mindset, you can get there. Thank you for watching and I'll see you in the next one. Thanks for watching the machine learning engineer road map for 2026. We hope that this video gave you a clear path to becoming a successful ML engineer. Ready to take the next step? Explore the Simply Learns professional certificate in AI and machine learning in partnership with Purdue University. Gain hands-on experience and industry recognized credentials to boost your career. Welcome to math refresher, probability and statistics. In this lesson, we are going to explain the concepts of statistics and probability. Describe conditional probability. Define the chain rule of probability. Discuss the measure of variance. Identify the types of gshian distribution. Basic of statistics and probability. Probability and statistics. Data science relies heavily on estimates and predictions. A significant portion of data science is made up of evaluations and forecast. Statistical methods are used to make estimates for further analysis. Probability theory is helpful for making predictions. Statistical methods are highly dependent on probability theory and all probability and statistics are dependent on data. Data is information acquired for reference or research via observations, facts, and measurements. Data is a set of facts structured in the form that computers can interpret such as numbers, words, estimations, and views. Importance of data. Data aids in seeing more about the information by identifying possible connections between two features. Data assists in the detection of distortion by uncovering hidden patterns based on prior information patterns. Data may be utilized to anticipate the future or predict the current state of affairs. Also, data aids in determining whether two pieces of information have any instance in common or not. Types of data. Data might be quantitative that is data that can be measured or counted in numbers or it may be qualitative which is data which is generally divided into groups or in simpler words which cannot be counted or measured in numbers. Let's consider an example. A customer information data of a bank may contain quantitative and qualitative data. Consider this snapshot where we have customer ID, surname, geography, gender, age, balance, has C or card is active member. Amongst these variables we can see surname is mostly qualitative as it cannot be counted and measured in numbers. Geography and gender are also qualitative as they cannot be counted in numbers and are mostly groups. has C or card that is has credit card and is active member although are containing numerical in form but these are categorical that means these have been divided into groups of one and zero that represent yes and no as an answer hence these two variables are also qualitative customer ID is again although a numerical data however the significance or intuition behind Customer ID is categorical. Hence, it may be kept in the qualitative data also. However, age and balance these are numerical information which have been measured or counted and numerical operations can be performed on them. Hence, these are under quantitative data categories. Introduction to descriptive statistics. Descriptive statistics. A descriptive measurement is summary measure that quantitatively portrays the most important features of a set of data allowing for a better comprehension of the information. Data can be measured as different levels. The levels of measurement describe the nature of information stored in the data assigned to the variables. Qualitative data can be measured as nominal or ordinal. Quantitative data can be measured in terms of interval and ratio type. Nominal data. The data is categorized using names, labels or qualities. For example, brand name, zip code, and gender. Ordinal data can be arranged in order or ranked and can be compared. Examples include grades, star reviews, position, and race, and date. Interval data is the data that is ordered and has meaningful differences between the data points. Example, temperature in Celsius and year of birth. Ratio data is similar to the interval level with the added property of inherent zero. Mathematical calculations can be performed on both interval as well as ratio data. For example, height, age, and weight. Population versus sample. Before analyzing the data, it's important to figure out if it's from a population or a sample. Population is a collection of all available items as well as each unit in our study. Sample is a subset of the population that contains only a few units of the population. Population data is used for study when the data pool is very small and can give all the required information. Samples are collected randomly and represent the entire population in the best possible way. Measures of central tendency. The central tendency is a single value that aids in the description of the data by determining its center position. Measures of central tendency are sometimes known as summary statistics or measures of central location. The most popular measurements of central tendency are mean, median, and mode. The normal distribution is a bell-shaped symmetrical distribution in which mean, median, and mode all are equal. The curve over here shows the bell-shaped curve or the normal distribution of variable X. The point over here that is X1 is the point which represents the mean, median and mode of this distribution. Mean mean is calculated by dividing these sum of all data values by the total number of data values. It gets affected when there are unusual or extreme values. It is sensitive to the outliers. Mean can be calculated as summation over all the values of X in a collection divided by the size of the collection. For example, we have a collection where we have values as 7 3 4 1 6 and 7. We find out the sum of these values which is 28 and there are total of six values. So 28 / 6 gives us a mean value of 4.66. Median, it is the middle value in the set of the data that has been sorted in ascending order. It is a better alternative to mean since it is less impacted by outliers and skewess. It is closer to the actual central value. Median is calculated differently for different sizes of data. Differentiated as if the total number of values is odd or if the total number of values is even. If the size of the data is odd. For example, in this case we have five elements. After sorting whatever middle value we get that means n + 1 by 2 term in this case 5 + 1 / 2 that is the third term which is four is the median value. In case when the total number of values is even like here there are six values. The average or the mean of the two central values is considered as the median. In this case the median is the mean of 6 and four which is five. Mode. Mode represents the most common value in the data set. It is not at all affected by extreme observations. It is the best measure of central tendency for highly skewed or non-normal distribution. Mode for categorical data is determined by estimating the frequencies for each categories and then the category with the highest frequency is considered to be mode. Like in this case seven has the highest frequency. Hence seven becomes the mode value. However, in case of continuous data or quantitative data, the calculation of mode is slightly different. The first step in calculation of mode is dividing the data into classes which are equal with then getting the frequency of data points lying in within that range of classes and finally selecting the class with the highest frequency. Using the range of that class and the frequencies, we can get the final mode Using the formula L plus FM minus F_sub_1 ultiplied to H / F minus F_sub_1 plus FM minus F_sub_2. Here L is the lower limit or the lower observation of the mode class. H is the size of the mode class. FM is the frequency of the mode class. F_sub_1 is the frequency of the class proceeding to mode and F_sub_2 is the frequency of the class succeeding to mode. This gives us the final mode Mean versus expectation. Now let's talk about mean versus expectation. So in general we use the expected value or expectation when we want to calculate the mean of a probability distribution that represents the average value we expect to occur before collecting any data. And mean on the other hand mean is basically used when we want to calculate the average value of a given sample. This represents the average value of raw data that we may have already collected. We can understand this by using a simple example. Now to calculate the expected value of this probability distribution, we can use a specific formula from the previous discussion. This is going to be the expected value where X is going to be the data value and this PX is the probability of value. For example, we could calculate the expected value for this probability distribution to be as shown. So here it will be 1.45 goals. So this represents the expected number of goals that the team will score in any given game. And then if you talk about calculating mean, so we typically calculate the mean after we have actually collected raw data. For example, suppose we record the number of goals that a soccer team will score in 15 different games. Now to calculate the mean number of goals scored per game, we can use the following formula where sum of x is basically the sum of all the goals divided by n and the number of records or we can say the sample size. It is as shown on the screen. So this represents the mean number of goals scored per game by the team. Measures of asymmetry. The difference between the three distinct curves can be studied in this image. The central curve is the normal or no skewess curve here. mean, median and mode all lie on the same point. This normal curve is symmetrical about its mean, median and mode. That means the left hand side of the curve is a mirror image of the right hand side of the curve. However, in case of negatively skewed data, the tail is elongated on the left hand side and the mean is smaller than the mode and the median values or is on the left hand side of the mode. Hence indicating that the outliers are in the negative direction. On the other hand, in case of positively skewed, the data is concentrated on the left hand side of the curve. While the tail is elongated or longer on the right hand side of the curve, the mean is greater than the mode and median or is on the right hand side of the mode and median indicating that the outliers are in the positive direction. Let's consider an example. The graph here shows the global income distribution for the year 2003 2013 and a projection for 2035. If we see the global income distribution statistics for 2003 it is highly right skewed. We can observe in the previous graph that in 2003 the mean of $3,451 was higher than the median of $1090. The global income is definitely not evenly distributed. The majority of people make less than $2,000 each year, while only a small percentage of the population earns more than $14,000. Measures of variability. Dispersion. The measure of central tendencies provide a single value that addresses the full worth. However, the central tendency cannot depict the viewpoint entirely. The metric of dispersion helps us focus on the inconsistency in the data spread. Measures of dispersion describe the spread of the data. The range, intercortile range, standard deviation and variance are examples of dispersion measures. Range. The range of distribution is the difference between the largest and the smallest amount of data. The range, for example, does not include all of a series positive aspects. It concentrates on the most shocking aspects and ignores that aren't considered critical. For example, for a set 13, 33, 45, 67, 70, the range is 57. That is the maximum of this which is 70 minus the minimum over here which is 13. Variance. Variance is the average of all squared deviations. It is defined as the sum of squared distance between each point and the mean or the dispersion around the mean. The standard deviation is used as variance suffers from a unit difference. Variance can be computed as sigma square summation over x - mu^ 2 divided by n where mu is the mean of the data, x is the individual data point and n is the size of the data. This representation is for a population for a sample data variance can be computed as x minus xar whole square summation over it divided by n minus one. Here xbar is the mean of these sample data and n is the sample size. The units of values and variance are not equal. So another variability measure is used. Standard deviation. Standard deviation is a statistical term used to measure the amount of variability or dispersion around a mean. The standard deviation is calculated as the square root of variance. It depicts the concentration of the data around the mean of the data set. Standard deviation as indicated previously can be computed as square root of variance for a population data. Standard deviation sigma can be computed as square root of summation over x i minus mu^ square / n where mu is the mean of the data x i are the data points and n is the size. Let's consider an example. Let's find out the mean, variance, and standard deviation for this data. The data values are three, 5, 6, 9, and 10. To find out the mean, we first find the sum of all these data values that is 33 and divide it by the count, which is five. We get the mean of 6.6. To compute the variance, we start by computing the deviation. That is X minus the mean of X. Here 3 is one of the values of the data and 6.6 is the mean. So 3 - 6.6 squared and we do that to find out sum of all the deviations divided by the count We end up getting an overall variance of 6.64. Standard deviation as we know is measured at square root of variance that is square<unk> of 6.64 which amounts to 2.576. Measures of relationship. Measures of relationship coariance. Coariance is the measure of joint variability of two variables. It measures the direction of the relationship between the variables. It determines if one variable will cause the other to alter in the same way. Coariance between variable X and Y can be computed as summation over the product of X I - Xar and Y I - Y bar the whole divided by N minus one. Here Xar and Y bar are the mean of X and Y respectively. The value of covariance can range from minus infinity to a plus infinity. Correlation. Correlation is normalized coariance. It measures the strength of association between two variables. The most common measure for correlation is the Pearson correlation coefficient. Correlation between two variables X and Y can be measured with respect to coariance as coariance between X and Y divided by the standard deviation of X and standard deviation of Y. The value of correlation ranges from a negative 1 to positive 1. Types of correlation. Correlation can be either a positive correlation, zero correlation or a negative correlation. The first picture over here represents a perfect positive correlation wherein a straight line with a positive slope is representing the relationship between the two variables. Zero correlation means that the line representing the relationship between the two variables is horizontal to the xaxis. Perfect negative correlation can be represented by a straight line with a negative slope. Correlation equals to 1 implies a positive relationship. That is when one variable increases the other variable also increases. A correlation value of negative 1 implies a negative relationship. That is when one variable increases the other decreases. The correlation coefficient of zero shows that the variables are completely independent of each other. Here we have two variables height and weight. To compute the correlation between height and weight, we use the correlation formula as coariance of X and Y divided by standard deviation of X and standard deviation of Y. Here height is the X variable and weight is the Y variable. First to compute coariance we compute the x - xar and y - y bar values and then the product of them. We then compute x - xr² and y - y bar square values to compute the standard deviations of height and weight respectively. Correlation as we know has been defined as covariance of x and i and y divided by standard deviations of x and y. This can also be represented as summation over x - xr multiplied to y - y bar divided by square root of summation over sum of squared deviations that is x - xr square multiplied to square root of summation over y - y bar square that is sum of square deviations for y. Now let's find out values to put into this formula. First we find out the overall sum of height to get the mean of height which is 5.14. Similarly we get the sum of weight to get the mean of weight as 50. We now get the summation over x - xr multiplied to y - y bar to get the numerator for the formula. Then we compute x - xr square summation and y - y bar square that is sum of squared deviation of x and y respectively. Now we put in the values in this final correlation formula to get a correlation value of 0.889. This indicates that height and weight have a positive relationship. It is evident that as height grows, weight also increases. In this module, we will be talking about expectation and variance. So the expected value or we can say mean of a given variable that we can denote by X is a discrete random variable where it is a weighted average of the possible values that X can take and each value is going to be according to the probability of that specific event occurring. So usually the expected value of X is denoted by a simple formula where we can define the expectation based on the X parameter. which is going to be the sum of each possible outcome multiplied by the probability of the outcome occurring. So in more concrete terms, the expectation is what we would expect the outcome of an experiment to be on average. We can take an example for the coin. If a coin is being tossed 10 times, then one is most likely to get five heads and five tails. Same logic can be discussed if we talk about another example of rolling a die. So there are six possible outcomes when you roll a dieice 1 2 3 4 5 6 and each of these has a probability of 1 by 6 of occurring. So we can say that the expectation is going to be 1 multiplied by the probability of that happening which is going to be 1x 6 + 2x 6 + 3x 6 + 4x 6 + 5x 6 + 6x 6 and that is going to give us 3.5 as an output. The expected value is 3.5. So if you think about it, 3.5 is halfway between the possible values that I can take and this is what we should have expected. Next we talk about the concept of variance. So variance of a random variable allows us to know something about the spread of the possible values of the variable. So for a discrete random variable X, the variances of X is going to be denoted by using a simple formula that is going to be var equals E X - M the whole square where M is basically the expected value of the expectation of X. So this is more like a standard deviation of X which can also be represented by using this formula. So the variance does not behave in the same way as expectation when we multiply and add constants to random variables. So now there are two different type of variance that we can have a fair understanding on. First of all we have low variance and then we have high variance. So low variance simply means that there is a small variation in the production of the target function with changes in the trading data set and at the same time high variance as we can see here high variance shows a large variation in prediction of the target function with changes in the trading data set. So a model that shows high variance learns a lot and perform well with the training data set and it does not generalize well with the unseen data set and that's why as a result such a model gives good results with training data set but shows high error rates on the test data set and since the high variance a model learns too much from the data set it leads to an overfitting of the model. So model with high variance will be having couple of issues like it may lead to overfitting or it may also lead to increase in model complexities. Next we have skewess. So skewess in simple terms is basically a measure of asymmetry of a distribution. So distribution is asymmetrical when its left and right sides are not the mirror images. Right now this is a mirrored image and a distribution can have right positive or we can say negative or it can have zero So right skewed in this scenario is basically the distribution is longer on the right side of its peak and a left skew distribution is going to be we can say where it is longer on the left side. So we can see we have this one as a part of right side. It is more elongated towards the right side and this one is more elongated towards the left side. So we can think of skewess in terms of tails. A tail is long tampering and the end of a distribution. So it simply indicates that they are observations at one end of the distribution but that they are relatively infrequent. So a right skew distribution has a long tail on the right side as you can see here. So the number supports observed. Let's say we have a data on a per year basis. So again we can have a more skewess towards the right side where data is being dropping as we continue to increase the number of years. For example we may have a high sales towards the beginning of year suppose in 2022 but again as we proceed to 2023 second half we are seeing the dip in performance. So that is rightly skewed and same way let's suppose if we started with the sales figure it was really less in suppose 2002 but again as we proceeded to 2023 now our sales have been gradually increasing so it's more like skew towards the left section as a part of negative skew. Next we have curtosis. So curtosis is basically a measure of the tailness of a distribution. So tailness is how often the outliers occur and acts as courtesis. Is the tailness of the distribution related to a normal distribution. So a distribution with medium curtsis is called as meocurtic. A distribution with low curtosis like this one. This is called as the platicurtic and then distribution with high curtosis like this one. This is called as the leptocortic. So tails here they are tapering ends on either side of a distribution like this. So they represent the probability or the frequency of values that are extremely high or extremely low to the mean. In other words, tails here represents how often the outliers occur. So there are three type of curtosis. We have platocurtic which is negative, leptoccuric which is a positive towards the upper end and then we have messertic which is a normal distribution. So messertic is the medium tail. So normal distributions they have a curtosis of three. So any distribution with a curtsis of a prox value of three is going to be messertic. And curtosis is described in terms of excess curtises which is curtosis minus3. And since normal distribution they have a curtosis of three axis curtises makes comparing a distribution curtosis to a normal distribution even easier. Introduction to probability. Probability theory. Probability is a measure of the likelihood that an event will occur. Let's consider an example of coin toss where the chances of getting heads on a coin are 1 by two or 50%. The probability of each given event is between 0 and 1 both inclusive. Sum of an events cumulative probability cannot be greater than one. Hence the probability of an event x lies between zero and one. This means that the integral of probability of distribution over X equals to 1. Conditional probability. Conditional probability of any event A is defined as the probability of occurrence of A given that event B has previously occurred. Condition probability of event A given B can be estimated as probability of A intersection B that is probability of both A and B happening together divided by the probability of B. It is also written as that probability of A intersection B equals to probability of A given B multiplied to probability of B. In a coin, we are doing a two coin flip. Coin one gets heads, tails, heads, and tails in subsequent flips. while coin two gets tails, heads, heads, and tails in the subsequent flips. Now, the probability that coin one will get a head is 2 out of four. While the probability that coin two will get heads is again two out of four. The probability that both coin one and coin two will have a heads is just one out of the four flips. Hence the probability that coin one will get heads given that coin 2 is already heads can be computed as probability of coin one edge intersection coin 2 edge that is 1x4 divided by probability of coin 2 edge that's a given that is 2x 4 which is going to be 0.5 or 50% based base theorem Base theorem calculates the conditional probability of an event based on its prior probabilities. Basically base theorem incorporates the prior probability distribution to predict the posterior probabilities base theorem for conditional probability can be expressed as probability of A given B equals probability of B given A divided by probability of B multiplied to probability of A. Base theorem allows updating the probability values by using new information or evidence. Here probability of A is known as prior probability. That is the probability of event before any new data is collected. Probability of A given B is known as the posterior probability. It is the revised probability of an event occurring after taking into consideration the new information probability of B given A is known as the likelihood and probability of B is probability of observing an evidence B model. An example consider an example for calculating the likelihood of having diabetes based on frequency of fast food consumption. Here is the observed data. Let's say the fast food audience is 20%. Diabetes prevalence is 10% and 5% is fast food and diabetes. The chances of diabetes given fast food that is the conditional probability of D given B can be calculated as probability of diabetes and fast food together divided by probability of fast food. That means 5% divided by 20%. that equals 25%. Define an analysis can state eating fast food increases the chance of having diabetes by 25%. The multiplication rule of probability if events A and B are statistically independent and probability of A intersection B can be given as probability of B. However, probability of A intersection B is also given as probability of A multiplied to probability of B. Here probability of A given B equals to probability of A when we assume that probability of B is non zero. Similarly, probability of B equals probability of B given A assuming probability of A is non zero. Chain rule of probability joint probability distributions over many random variables can be reduced into conditional distributions over a single variable. It can be expressed as probability of X1 X2 so on until Xn equals probability of X1 intersection probability of X I given probability of X1 till X I minus one. For example, the joint probability of A, B and C can be given as probability of A given B. C multiplied to probability of B given C multiply to probability of C. Logistic sigmoid. The logistics function is a type of sigmoid function that aims to predict the class to which a particular sample belongs. Its outcome is discrete binary value. a probability between zero and one. The logistic sigmoid is a useful function that follows the yes curve. It saturates when the input is very large or very small. Logistic sigmoid is expressed as sigma of x= 1 upon 1 + e to the power minus x. The logistic sigmoid can be expressed as sigmoid function of x is given as 1 upon 1 + e ^ min - x where e is the ooler's number. Gshian distribution. The gossian distribution is a type of distribution in which data tends to cluster around a central value with little or no bias to the left or right. It is often referred to as normal In absence of prior information, the normal distribution is frequently a fair assumption in machine learning equation. The formula for calculating Gaussian distribution is described as the normal distribution of X. That is the function of X given mean as mu and variance is sigma square can be calculated as 1 upon sigma square<unk> of 2<unk>i. E to the power minus/ X - mood divided by sigma square where mu is the mean or peak value which also is the expected value of X. Sigma is the standard deviation. Sigma square is the variance. A standard normal distribution has a mean of zero and a standard deviation of one. Gshian distribution can be univariate which describes the distribution of a single variable X. It can also be multivariate where it can just use to describe the distribution of several variables. It is represented in 3D of ND formats. Law of large numbers. Now let's talk about law of large numbers. The law of large numbers states that an observed sample average from a large sample will be close to the true population average and that it will get closer in the larger sample. So the law of large number does not guarantee that a given sample spatially a small sample will reflect the true population characteristics or that a sample does not reflect the true population will be balanced by a subsequent sample. This is for the law of large numbers to express the relationship between scale and growth rate. So there are multiple examples through which we can understand and it is widely used in statistical analysis in working with the central limit theorem in terms of the business growth. So there are multiple real time setup in which these are going to be used. So if you talk about tossing a coin, so tossing a coin in a number of times will give us two different type of outcomes. The result will spread evenly between head and tails and the expected average value is going to be half. That means 50 * tails and 30 * heads. But again, if you toss a coin 1,00 times, then the result can be in different manners because out of 1,000, let's say 850 times it has been head and only 150 times it has been tails and so on. So that's why the possibility of one event occurring is going to be changed in large sample sets as compared to a small sample sets as in let's say 10 times. So the number of heads and tails unbalanced for lower number of trials. So we can see it is unbalanced. But again as soon as we toss more number of coins more leans towards the balance value or we can see the observed averages. Next we have P value. So p value is basically a number calculated from the statistical test that describes how likely we are to have found a particular set of observations if the null hypothesis were true. So p values are used in hypothesis testing to help decide whether to reject the null hypothesis. And the smaller the p value, the more likely we are to reject the null hypothesis. So we have a term called as null hypothesis. So all statistical tests they have null hypothesis. So for most tests the null hypothesis is that there is no relationship between our variables of in first or that there is no difference among groups. For example in a two-tail t test the non-hypothesis is that the difference between two groups is going to be zero. So p value is going to tell us how likely it is that our data could have occurred under the null hypothesis. It is done by calculating the likelihood of a test statistic which is the number calculated by a statistical test using our data. So p value tell us how often we would expect to see a test statistic as extreme or more extreme than one calculated by a statistical test. if the null hypothesis of the test was true. So there are multiple limitations as well. So first one is the results can be significant but again they are they may not be practical as we have compared it can be based on multiple hypothesis for a game for the healthcare test. If the test is going to be positive or not it may show even values of the effect of a variable but not the magnitude in real life. What exactly is going to be the application of a drug test being failed in pharma company? Therefore, it is recommended to use confidence and levels in addition to the p values to quantify or we can say to give a solid figure to the reserve which we are going to get. The p values they are interpreted as supporting or we can say refuting the alternative hypothesis. So p value can only tell you whether or not the null hypothesis is supported. It cannot tell us whether our alternative hypothesis is true or why. So the risk of rejecting the null hypothesis is often higher than the p value. So especially when we are looking at a single study or when using small sample sizes. So this is because the smaller frame of reference, the greater are the chance that as we stumble across a statistically significant pattern completely by accident. Key takeaways. Key takeaways. Probability and statistics structure the premise of the data. The data helps in anticipating the future or gauging in view of the past patterns of information. that helps to describe the data by identifying these central positions. The mean, median, and mode are the measures of central tendencies. The distribution where the data tends to be around a central value with a lack of bias or minimal bias towards the left or right is called as gshian distribution. My name is Richard Kersner with the simply learn team. That's get certified, get ahead. We're going to cover mathematics for machine learning. So today's agenda is going to cover data and its types. Then we're going to dive into linear algebra and its concepts, calculus, statistics for machine learning, probability for machine learning, hands-on demos, and of course throwing in there in the middle is going to be your matrixes and a few other things to go along with all this. Data and its types. Data denotes the individual pieces of factual information collected from various sources. It is stored, processed, and later used for analysis. And so we see here uh just a huge grouping of information, a lot of tech stuff, money, dollar signs, numbers uh and then you have your performing analytics to drive insights and hopefully you have a nice share your shareholders gathered at the meeting and you're able to explain it in something they can understand. So we talk about datas types of data we have in our types of data we have a qualitative categorical you think nominal or ordinal and then you have your quantitative or numerical which is discrete or continuous and let's look a little closer at those data type vocabulary always people's favorite is the vocabulary words okay not mine uh but let's dive into this what we mean by nominal nominal they are used to label various just uh label our variables without providing any measurable value. Uh country, gender, race, hair, color, etc. It's something that you either mark true or false. This is a label. It's on or off. Either they have a red hat on or they do not. Uh so a lot of times when you're thinking nominal data labels, uh think of it as a true false kind of setup. And we look at ordinal. This is categorical data with a set order or a scale to it. Uh and you can think of salary range is a great one. Uh movie ratings etc. You see here the salary range if you have 10,000 to 20,000 number of employees earning that rate is 150 20,000 to 30,000 100 and so forth. Some of the terms you'll hear is bucket. Uh this is where you have 10 different buckets and you want to separate it into something that makes sense into those 10 buckets. And so when we start talking about ordinal, a lot of times when you get down to the…

Transcript truncated. Watch the full video for the complete content.