News

Top 18 Free Training Resources for AI and Machine Learning Skills (Plus 3 Great Paid Ones, Too)

From books to training courses to datasets to toolkits, here are some great, no-cost resources that will help you transform your current programming skills to meet the AI and machine learning needs of the future.

There are literally hundreds, if not thousands, of fantastic and free AI/machine learning resources out there for developers hungry to improve their skills. So where do you start?

We scoured the forums and message boards, asked working machine learning and AI developers, and consulted with researchers and other experts to come up with the following list that can get you, the working developer, going in the right direction.

We're starting with one of the oldest resources but also one of the most widely regarded: Published in 1992, the book is subtitled "Case Studies in Common Lisp," so it isn't written for a language widely popular for AI today (although Lisp is still used in some situations). Even so, no matter what language you're programming in, many swear by this classic resource because of the way author Peter Norvig (now a director of research at Google) eruditely yet clearly and approachably covers all aspects of AI programming. You can get the book (and all the code ) for free from the GitHub link above, or you can also purchase it in paperback, digital or Kindle form from Amazon .

From one of the oldest resources to one of the newest, Google's Machine Learning Crash Course (which focuses on using TensorFlow, the company's open source machine learning framework) became an instant hit when it was released earlier in 2018. The roughly 15-hour, free course consists of 25 lessons, 40 exercises, video lectures from Google researchers and other interactive elements. When you're done, step over to Google's AI page for even more free offerings.

It's hard to find a credible list recommending AI or machine learning resources where this course isn't near or at the top, as even working AI professionals refer to it as "the gold standard" of machine learning education. Taught by Stanford Assistant Professor Andrew Ng , it consists of 11 weeks of training and is not by any means easy -- your skills will be tested. Even so, it's free: You can choose to pay $79 to earn the certificate or not pay and only walk away with the knowledge you gained from the course. If you're serious about getting into machine learning as a career, get your background skills prepared using other resources on this list and then sign up for this course. (Bonus: Don't miss Ng's new book, Machine Learning Yearning .)

If you are going to be using one of the most popular languages for machine learning -- Python, and specifically Python 3 -- then you may want to start with Zed Shaw's excellent yet painful "Learn Python 3 the Hard Way: A Very Simple Introduction to the Terrifyingly Beautiful World of Computers and Code." It's also available for purchase and comes with bonus videos and exercises from the author when you buy it from him. While this book doesn't focus on AI, it will get you to brass tacks on Python by the time you're done -- painfully, by all accounts (Shaw's not going to hold your hand, and that's the entire point).

IBM offers a wealth of resources for developers who want to get started with Watson , its ubiquitous cognitive computing tool/AI platform for the enterprise (it even offers a free IBM cloud account to go along with it). And once you get everything set up, the company offers various Starter Kits that take you step-by-step through different scenarios that, while simple, show you a bit of what Watson can offer enterprises and how it accomplishes them. You can then dive in deeper if you like what you see, but these starter kits are not only a great introduction to Watson, but to the power of AI in general (and they're fun, too). This is a great place to start if you want to begin by getting your hands dirty.

6. Kaggle.com (Datasets)

To actually get hands-on with machine learning, you need data. Kaggle is a probably the most popular of many Web sites within the data science world offering free datasets. From U.K. traffic accidents to top Spotify tracks to stock data to the air quality in India, you're bound to find something you'll want to work with (and it correlates datasets from many other sites, so you're covering a lot of ground using Kaggle). Once you do get your feet wet, pop over to the site's competition page, which lists a wide variety of machine learning competitions, with prizes currently ranging everywhere from "swag" to $100,000. And don't miss the site's starter challenge.

Bonus: 3 Great Paid AI and Machine Learning Resources

Not everything in life is free. Here are three other great resources we think developers will want to know about:

1. Python for Data Science and Machine Learning Bootcamp (Online Course)
This popular (particularly for beginners) Udemy course from instructor Jose Portilla includes 22.5 hours of video, nine articles, several resources and a certificate of completion. And while the list price of the course on Udemy is $194.99, with the various promotions and discounts that Udemy almost always runs, your purchase price will most likely be a lot lower. For example, when this author was looking at the course, Udemy was offering a special where all courses -- including this one -- were being offered for $12.99.

2. AI Live!, Part of Live! 360 (Live Conference)
Yes, this is a plug for a sister event for this site, but Live 360! is a very popular live training conference held every year in Orlando, Fla., and now AI Live! is part of it. Note that if you're planning to attend Live! 360/Visual Studio Live! in Orlando anyway, AI! Live is kind of free in that Live! 360 has the "6 events, 1 price" scenario -- you don't have to pay for it separately as all the sessions are included in your overall Live! 360 registration fee.

3. Hands-On Machine Learning with Scikit-Learn and TensorFlow (Book)
This book by Aurélien Géron, an AI engineer, former Googler and current head of the startup Kiwisoft, is not only an Amazon best-seller, but also regularly ends up on "must-read" recommendation threads (like this one) that appear across the Internet. You may also want to follow Géron on LinkedIn, as that's where he posts updates of the latest language versions of his books, new videos and more.

7. MIT Linear Algebra Class (Online Training)

While math is important for all areas of programming (and vice versa), you may want to brush up on your linear algebra in particular before you dive into certain aspects of machine learning. If you need more than just a quick refresher (here's an excellent one), the online course from MIT OpenCourseWare linked above offers the entire curriculum online for free, from lectures to exams to study materials.

This book is available free in .PDF format via the link above, and the site offers links to all the lab code. Written by professors at USC, Stanford and the University of Washington and focused on R -- the language of statistical computing that is often used for machine learning and AI programs in this area -- the book has been described as "the 'how to' manual for statistical learning." Once you're done with this book, move on to the authors' follow-up, " The Elements of Statistical Learning ," also available for free online (although both can be purchased, as well).

Google isn't the only vendor to make its AI training course free this year; in April, Microsoft announced it was making its internal AI course free and available to everyone via the Internet. The extensive learning track consists of 10 individual courses (one of which is a final project), with titles like "Introduction to Python for Data Science," "Build Machine Learning Models" and "Develop Applied AI Solutions." Each course lasts a certain number of weeks and has a separate paid certificate option (view a sample one here). You can take one course, a mix of courses or all of them to earn the overall program certificate.

10. Google Dataset Search (Datasets)

Google knows how much machine learning pros and students need (and love) data. They also have to be able to find what they need in the right format easily and quickly. So the company (surprisingly only just recently) has launched a search option where you can search the Internet for only datasets (direct link above). The best part is that the results (at least, when they come from a well formatted site like Kaggle) include the information you really need, like data format and when the set was last updated. Google also notes the number of scholarly articles that cite the resulting dataset, if any, and provides descriptive details so you can scan through your options fairly quickly instead of jumping back-and-forth to destination pages.

Yes, we've included a second MIT OpenCourseWare course here, but start watching the videos linked above and you'll see why -- this course covers so much and it covers it well. Even if you don't want to take the entire class, you can cherry-pick your way through the video list and watch only the ones whose topics are of interest to you at that particular point in your AI journey.

This online book, written by Y Combinator Research Fellow Michael Nielsen , covers everything from how backpropagation works to issues with training neural networks and differences between shallow and deep networks -- along with plenty of hands-on exercises. And while it's all completely free and accessible to everyone, the author does suggest that those who enjoy the book's contents make a small donation ($5 recommended) via the book's Web site.

When you're ready to really dive down into neural networks, jump over to YouTube and dive into this excellent collection of videos created by the Université de Sherbrooke's Hugo Larochelle. Many of the comments under the video are helpful, too, but what a lot of those commenters don't realize is that the series doesn't just consist of the videos -- you can get .PDFs of the slides, recommended readings from the author and more in an excellent outline of the entire course here.

This completely free, open source, commercial-grade toolkit from Microsoft offers deep-learning capabilities for unstructured data and lets users "easily realize and combine popular model types such as feed-forward DNNs, convolutional neural networks (CNNs) and recurrent neural networks (RNNs/LSTMs). CNTK implements stochastic gradient descent (SGD, error backpropagation) learning with automatic differentiation and parallelization across multiple GPUs and servers." So basically, when you're ready to start playing around in the area of deep learning, here's one place you may want to begin. Check out the model gallery for more of what you can do with it.

Once you are ready for machine learning how-tos, this is one article series not to miss. Written by Microsoft Research's James McCaffrey and focused on both R and Python, the series walks readers through topics such as "Linear Regression with R," "Neural Networks Using the R nnet Package," "Neural Network Back-Propagation Using Python," "Neural Network L1 Regularization Using Python," plus so much more. And it's an ongoing series, so new content is added every month.

When it comes to "getting" machine learning, you need to understand both the theoretical and the practical. This 2014 book from Cambridge University Press is great not only because it's well-respected and available for free in .PDF format from the link above, but also because you can access a number of courses that use the book online here , supplementing your own learning in whatever way you choose.

This site is exactly what you'd expect it to be from its name: a neural network setup device that you can play with right in your browser. (The site's tagline is "Don't Worry. You Can't Break It. We Promise.") There's a wide variety of factors you can adjust; you can add numerous hidden layers and even turn on/off the various controls further down on the page (and then just refresh). It's also available via GitHub in case you've thought of a whole new way you'd like to use it.

18. Open Data on AWS (Datasets)

Amazon Web Services (AWS) offers a fine collection of datasets, including satellite imagery, air quality data, rice genomes and brain scans. But the feather in its cap is really the Human Microbiome Project . More than 80 organizations and 300 scientists came together to gather this data, in which microbial samples were taken from 300 adult subjects, and then the microbes found were sequenced. The results generated 45TB of data over 1,128 referenced genomes, plus 2,400 "whole metagenome sequence datasets" from healthy subjects. The data can also be accessed on the project's portal site here .

About the Author

Becky Nagel serves as vice president of AI for 1105 Media specializing in developing media, events and training for companies around AI and generative AI technology. She also regularly writes and reports on AI news, and is the founding editor of PureAI.com. She's the author of "ChatGPT Prompt 101 Guide for Business Users" and other popular AI resources with a real-world business perspective. She regularly speaks, writes and develops content around AI, generative AI and other business tech. Find her on X/Twitter @beckynagel.

Featured