Get Started with Machine Learning and AI in 2023

hey youtube today we're going to be talking about how to get started in machine learning and ai let me start by saying congratulations because if you want to get started in machine learning that means you're about to enter one of the most exciting and rewarding fields that you can nowadays about six years ago i decided to make a career change into data science and machine learning and since then i've been able to work in various industries including hospitality energy pharmaceutical research and sports analytics machine learning is an evolving field and you'll always find yourself being challenged to learn more and new things that being said i want to walk through some of the things that i think you need to understand before starting your data science and machine learning journey so the first thing right off the bat is in order to do machine learning you're going to need to learn a programming language if you don't already for me this is an easy one because if you're starting your machine learning journey right now there really is one choice and that's python while there are other great programming languages out there some even designed for data science like r if you really want to get into machine learning python is the main language that you want to learn not only is python a great programming language to learn but there are also a variety of packages that are designed to do machine learning with python all the way from numpy which works with vector arrays to pandas where you can work with tabular data to scikit-learn pi torch and tensorflow which are some of the fundamental machine learning libraries when i decided to transition into data science the first course i took was mit's introduction to programming with python you can audit this course for free through edx and it's taught by some of the best professors in the world i highly recommend it if you have no programming background so either while you're learning about python or if you're already familiar with coding in python the next thing you want to learn is a good fundamentals in math and statistics the depth you need to learn each topic really depends on what you plan to do with machine learning but at least a good foundational understanding of linear algebra some optimization and then also statistics and probability hopefully you have some foundational understanding of mathematics but you can always brush up on them by searching for good youtube videos or i would suggest just going through some khan academy courses on the basics of linear algebra and statistics one thing i will say about math and statistics is not to get too bogged down in the details as you're doing machine learning and practice you're going to find yourself going back and relearning some of these topics but they're always going to be there at your fingertips there's no real reason to try to memorize every single math and statistic concept out of the gate at this point you're familiar with python you understand the syntax you also understand some basics about math and statistics and you just need to get your hands dirty messing with some data sets numpy and pandas are a must know for anyone learning machine learning or data science and i would suggest watching some tutorial videos i have a few that you can check out it'll be really important that you know how to load in data read it in python and do some simple exploration even though data wrangling might seem like it's not part of machine learning you'll find very soon that it's essential tool that you'll need to be able to be comfortable with before you start actually creating machine learning models the fourth step i would suggest is getting familiar with the basic machine learning algorithms a great way to do this is just to walk through all the different machine learning models that are provided in the package scikit learn cyclic learn is open source and includes algorithms for clustering classification regression problems and these are all things that you need to have a good foundational understanding of if anything to know when you see a certain data set what type of algorithms might be a first thing to try there are a lot of tutorials out there about scikit-learn but the documentation is really good itself and if you find a data set that you're already excited about let's say you're excited about sports data or video game data find a data set about that and try to do some modeling based on that data set using scikit-learn you'll find that it'll be fun and you'll learn the different methods in the process at this point you might want to decide which area of machine learning to focus there's no official way that machine learning is broken up into but i like to think of it in three main groups the first is working with tabular or structured data this is the most common type of machine learning that you'll see in production nowadays it might be the type of data that you'd see in a spreadsheet or data about customers stock market and trend data this also would include a lot of time series forecasting and working with structured data requires a certain type of skill also the machine learning types that you'll probably be using in this application are more gradient boosted trees or linear and logistic regression now the second area machine learning that you may want to focus on is on unstructured data so this would be something like images video audio or signal data and typically the type of models that are used in those types of applications rely on deep learning this is because it's really hard to actually generate the features manually that you'd feed into a model and instead it's common to use deep learning models that identify these features for you this is also a really exciting field because there's a lot of growth at the same time i'd say deep learning applications are quite rare so if you're hoping to transition to a data science role sometime soon it's very hard and competitive to get a job where you're actually doing deep learning in practice the third group i talk about is just a catch-all for all the new fields that are coming out in machine learning specifically reinforcement learning there's been work with gans or generative models and there are also very specific types of machine learning that you might want to focus in on like graph neural networks or 3d neural networks now in terms of tools you'd need to know for each of these fields of machine learning i'd say if you're starting with structured data to focus on gradient boosted trees random forests linear and logistic regression some of my favorite package for this are light gbm xg boost and then of course scikit-learn for deep learning applications you're going to either pick tensorflow or pi torch as the main library that you start with tensorflow seems to be a lot of people's first choice because it's easier and more approachable to get started with and pi torch might be a little bit harder to understand at first but i've found that it becomes more customizable and there's a lot of details about the structure of the models that you create that you can configure a little bit easier but really pi torch or tensorflow if you pick one of the two you're going to be fine and then for things like reinforcement learning you're probably going to want to take a specialized course on reinforcement learning there's a lot of good books out there on the topic sutton and barto's introduction it's a really great starting guide and there are a lot of resources that you can find associated with it so the fifth step that i would suggest taking is to actually get your hands dirty and work on some real machine learning problems and in my opinion the best way to do that is through kaggle kaggle is an online community where you can share code learn about different algorithms and compete in live competitions if you find a good live competitions that that's currently active you can dive into it test out your machine learning skills alongside other people and usually the forums are a great place to ask questions and to learn as you go i've found that learning the foundations in machine learning is one thing but then applying it and actually doing it in practice is the best way to actually hone your skills so definitely even early on create a kaggle account start browsing what competitions are going on and reading some of the discussion forums because just doing that you'll learn a lot very last thing i'll say is just to remember that it's a marathon not a sprint so as long as you're learning something even if it's really small every day you're gonna slowly get to your end state and if you look back after a few years you're going to be surprised in all the things you learned the key is just to be consistent and learning every day and pushing yourself to learn something new and honestly that's why i love machine learning and data science because there's always something new to be learned alright thanks again for watching this video if you enjoyed it please like and subscribe and i'll see you next time