Learning How Machines Learn

Even a cursory observation of contemporary business and scientific publications would reveal a zeal for advances in artificial intelligence, machine learning, and deep learning. The formal field of AI research began decades ago in 1956, but recent progress and nascent practical applications have contributed to a growing sense that more widespread adoption is imminent.

Similar prognostications preceded earlier "AI winters" as well, but productive uses of machine learning and deep learning breakthroughs in business indicate that this time may actually be different. That is not to say that the goal of strong AI or artificial general intelligence will be realized in the near future. Only that the latest advances are stirring excitement and investment across a broad range of organizations.

Personally, the sub-field of deep learning is a fascinating area of study that combines my interests in learning, problem-solving, technology, and human behavior. Understanding how the brain assimilates and organizes information provides insight into both neocortical development in humans and approaches to computational deep learning. In a sense, trying to simulate learning, intelligence, and creativity in machines is to discover more about what it means to be human. This quest for knowledge is both enlightening and inspiring.

How To Begin Learning

Your long-term success at this endeavor will likely depend more on understanding why you are learning than which specific resources you access. Take some time to really think about what motivates you. It is this motivation that will carry you through the inevitable trough when it gets tough and quitting becomes an attractive option. For some, it might be a specific problem you wish to solve. For others, it could be the prospect of an engaging career. My inspiration comes from contributing to increasing creativity and complexity in the world.

After you have identified your Why? the focus shifts to what to do next. While researching the best learning resources I came across this Quora post titled simply How do I learn machine learning?. By now it boasts more than 100 answers and has been viewed over 1.8 million times. A complete review of all the responses is impractical but there were a number of resources and personal insights from credible sources that proved to be valuable inputs for constructing my learning strategy. I would encourage you to spend a few minutes reviewing the upvoted responses.

Recent graduates of university programs with a concentration in mathematics and computer science can skip ahead. For the rest of us, as the gulf widens between the present and courses on statistics, calculus, and algorithms, many will find it necessary to refresh their understanding of the theoretical underpinnings of the domain. If you count yourself among this group the list of resources below should prove useful. I personally learn better through video-based courses and hands-on practice, but I've listed several books for those who prefer this medium.

Online Courses:

Books:

Moving On To Advanced Topics

Once you feel comfortable with the basics it is time to evaluate machine learning and deep learning educational resources. Andrew Ng's Machine Learning course is one of the most frequently recommended online options. At present, you can gain full access to the course materials through Coursera for free.

Online Courses:

Books:

Personalize Your Approach

Each one of us has a unique educational background starting from early childhood and continuing through many years of informal or formal instruction. Consequently, there is no universal approach to understanding how machines and humans learn. Neither is a trivial subject to master, so prepare yourself now to stay engaged and enjoy the journey. Tailor your approach to align with your preferred style of learning and intrinsic motivators. Periodically remind yourself of why you embarked on this adventure. Surround yourself with like-minded people through online communities (Quora, DataTau, or reddit) and local meetups.

Most importantly you must design your practice to improve your performance and include some form of immediate feedback. Take into account your current skill level when choosing what to practice so that the material is neither too easy nor unattainable. Repetition of tasks you find comfortable will result in a learning plateau, so steadily ratchet up the difficulty to just beyond your comfort zone. Repeatedly practice in a manner that provides feedback until you get the hang of it before moving on to more challenging terrain.