Source: Deep Learning on Medium
PROVIDING DEFINITIVE ANSWERS
How do I actually learn more about machine learning?
If you’re overwhelmed by the resources out there, this is for you
Machine learning is a popular buzzword that has been circulating the net quite a lot in recent years, for a variety of reasons. Currently, deep learning is at the forefront of this field and the quality of research being conducted improves daily. As a result of the wealth of research, stories and information around the term bring produced, those who wish to learn more may feel overwhelmed. This article is meant to be a guide of sorts to aid the learning process of those inquisitive individuals and you the reader.
Before we divulge into resources that are most suited, we will first address the different ways you can decide to take this journey. Depending on your reason for choosing to enter this field, you can take a top-down or bottom-up approach. A top-down approach is when you start with the big picture of what you are trying to learn. In the case of machine learning, this may be online courses that go through the basics of supervised, unsupervised and reinforcement learning. Subsequently, teaching how to complete basic projects in each respect arm of machine learning. Whereas, a bottom-up approach is essentially breaking down the problem, or subject, into many smaller pieces. This means that rather than looking broadly, you focusing on conquering small sections individually before moving on, and this helps to build an in-depth knowledge of the subject. One may achieve this by deciding to tackle reinforcement learning first and learn how to create reinforcement learning algorithms that employ deep deterministic policy gradients.
Now the approach best suited to you will depend on what you are trying to learn, and why you intend to learn it. For an employee that has been tasked with creating a machine learning product, they may choose a bottom-up approach and research in-depth on the particular part of machine learning needed for that project. In comparison, a hobbyist may choose the top-down approach as they have more time on their hands and wish to explore the entirety of the subject. In addition, depending on your learning style, you may prefer certain resources over another as it increases the efficiency of your knowledge assimilation. Personally, I have found books to be helpful, but I can also really appreciate the interactive nature of certain online courses and how other features like forums allow for discussion and distribution of knowledge. Below I have listed the main types of resources that I feel would be beneficial to get you started, as well as a number of actual examples you can explore straight away.
Keep in mind that the list below is not exhaustive. If you feel that none really suit the information or learning approach you are looking for, Google, technology forums, and even YouTube reviews of resources are only a couple of clicks away.
Books are one of the best-known resources that people can use when trying to learn something new. Ranging from things like cookbooks, ‘An introduction to …’ to ‘how to … for dummies’. They are especially useful for machine learning due to the way they present new concepts in an informative, slightly more formalised manner. As with any other type of resource, there is an abundant amount of them so I would advise you to read the description or blurb of the book. This will allow you to understand what the book is about and whether or not it will be useful for enhancing your knowledge of machine learning. Tip: When possible, I would advise purchasing the online copy, this allows for an easy transition to trying out programming techniques or following links and software that may be mentioned.
Not everyone is a fan of books, and people may prefer learning digitally. You are able to start and stop as you like, and cover as little or as much of the course as you want. Online courses differ in the way they present content, length, and difficulty. Have a look at the ones below for a start and select one you think will be suitable for what you are looking for.
YouTube is a widely used platform and covers a wide range of topics. It can also be utilised to help us learn as machine learning professionals, or even corporations, produce content that is freely available for our consumption. Make sure you take advantage of this!
Publications & Articles
Articles offer another avenue through which to learn. I find them useful in approaching content from a different angle and introducing me to new topics. Generally, people are well informed about the topic they are writing about and this translates to educative articles. Subscribing to publications allows keeps you in the loop and is a good way to refresh your knowledge over time. Note: These are just a handful of many possible publications out there if you find these inadequate a little bit of research will provide you with many more.
Those who are still relatively new to the field may wish to gain a more in-depth understanding before attempting to read research papers. A good starting point, however, is the Two Minute Papers channel on YouTube. They create summary like videos of particularly interesting papers, making them more accessible and understandable. For the more “advanced” learners who want to be at the forefront of the field or look into and possibly experiment with recent developments themselves, the links below will be of use.
Podcasts are an extremely useful tool as they help to broaden your understanding of topics you may have come across. For example, you may listen to a podcast on how GANs are being used in research after learning about it from a video or online course. They also give you an insight into what is currently happening in the world of machine learning, and important projects people are undertaking. The podcasts listed below vary from short to longer episodes, which give you updates on interesting events and discussions to take note of.
“A learning curve is essential to growth.” – Tammy Bjelland
– Likewise, learning is a journey and one very rarely can call themselves a master of a subject. Especially with machine learning, a field that is expanding every day. There is simply too much information being produced. You will encounter learning curves, but push through and maintain a positive attitude.
– Try not to confine yourself to one resource. Switching up between books, interactive online courses, and even podcasts allows your knowledge to become more comprehensive. It enhances the likelihood that you will truly understand the knowledge you come across.