Deep learning has taken the world by storm in recent years, and for a good reason – it is an incredibly powerful tool for teaching machines to perform complex tasks.
However, deep learning as a subset of machine learning can be a daunting topic for some people.
Many newcomers to the field wonder if they need to dive deep into machine learning and understand a breadth of knowledge before getting their feet wet with deep learning.
This blog post will give my honest opinions on whether you can learn deep learning without first learning machine learning.
(it may be different from what you expect!)
What is Deep Learning?
Deep learning is a subset of machine learning consisting of models that learn trends and patterns in data the same way as traditional machine learning models.
However, due to how these models are constructed, you won’t know how these values or trends are derived.
You’ll often hear Deep learning models referenced as “black box” methods, as you’ll only understand the inputs and outputs and have no idea what goes on between the neurons in your neural networks. (Source)
A deep learning model with only one layer is equivalent to a model like linear regression or logistic regression.
This can be seen in the image below, where layer1 acts as our linear regression model.
Deep learning models are often composed of many layers, which makes them capable of learning complex patterns in data.
The term “deep” refers to the number of hidden layers in the model.
In the image below, we have 3 “deep” layers.
Deep learning is a powerful tool for solving problems such as computer vision and natural language processing.
Why You Should Learn Machine Learning Before Deep Learning
Just like you wouldn’t start learning to drive a car by jumping into a Ferrari, you wouldn’t start learning machine learning by trying to build the most complex models right away.
It’s important to start simple, learn the basics, and gradually increase the sophistication of the models, code, and problems you solve as you gain experience.
This will help you develop a strong understanding of the principles involved and prevent you from getting overwhelmed.
Once you have a solid foundation, you can start experimenting with more advanced techniques and see how they can be used to improve your models.
Also, very few machine learning problems need a deep learning solution.
You’ll be shocked at how accurate gradient-boosted trees are, often better than any deep-learning models you could build.
So, when starting out, don’t be discouraged if your models are relatively simple.
Just keep building on your knowledge, and eventually, as the problems you take on increase in complexity, deep learning solutions will be a natural next step for your repertoire.
While you “should” learn machine learning before jumping into deep learning – it doesn’t always mean you will.
(I will explain more in the next paragraph).
Reference:
https://www.displayr.com/gradient-boosting-the-coolest-kid-on-the-machine-learning-block/
Does Deep Learning Require Machine Learning Knowledge?
While I’m sure everyone on the internet will tell you that you need to learn this and that before you jump into deep learning, I’m going to go on a limb here and say that deep learning does not require a breadth of machine learning knowledge to start.
It ultimately depends on what you want out of your machine-learning knowledge and your machine-learning path.
For example, If all you are interested in is computer vision work, it doesn’t make a ton of sense to spend hours learning tree models (which dominate supervised learning).
Sure, it’s helpful to have a general understanding of the different types of machine learning algorithms, but if you’re solely focused on some subset of deep learning, you don’t need to worry about mastering every other aspect of machine learning.
Why not spend that extra time learning another deep-learning framework like PyTorch or TensorFlow?
Or spend time learning another CNN architecture? (the models commonly used in computer vision).
Just learn enough of the general terminology to get by and focus your energies on your subset of deep learning.
Now, if you’re after a more general role or career, like a machine learning engineer or data scientist, you’ll need to learn the basics of machine learning first – as they’ll dominate your career much more than deep learning.
Is Deep Learning Harder Than Machine Learning
Deep learning used to be much scarier than traditional machine learning; it was harder to implement, the hardware couldn’t handle it, and there was less support in terms of libraries and tools.
Also, the lack of forum support (like StackOverflow) was a huge time sink when building deep learning models.
Most of the problems you ran into, you had to solve yourself!
However, over the past few years, that has changed dramatically.
There are excellent deep learning frameworks like PyTorch and TensorFlow, and a thriving ecosystem of tools, forums, libraries, and GPU costs have plummeted.
As a result, deep learning is not nearly as scary as it once was.
It can be a lot of fun.
You’ll never “have enough” knowledge to justify learning the next thing.
Anyone can learn deep learning, and anyone telling you that you can’t learn it has a vested interest in keeping the number of people who know deep learning low.
So if you’re interested in exploring this exciting field, don’t be afraid to jump in.
Other Articles In Our “Without” Series
Here at Enjoymachinelearning.com, we’re exploring how things are interconnected. We have other articles in our “Without” series where we explore the technical differences between some of the common jargon within the data science and machine learning realm.
A few more of those articles are here:
- Artificial Intelligence Without Machine Learning
- Data Science Without Machine Learning
- Face Recognition Without Machine Learning
Frequently Asked Questions
While we wish we could respond to every email, we thought we’d answer the commonly asked questions here in this post.
Do I Need Coding To Learn Deep Learning?
You need to know how to code to learn deep learning. Manipulating data is not easy; without some programming knowledge, you’ll struggle to even get to the input stage of deep learning.
Is Deep Learning A Part of Artificial Intelligence?
Deep learning is a subset of machine learning, a part of artificial intelligence. Artificial intelligence is a vast field encompassing deep learning and standard machine learning (and much more).
Do Machine Learning Engineers Use Deep Learning?
Machine learning engineers use deep learning, though most machine learning problems are solved without it. Deep learning solutions need tons of data, and that amount of data is usually a luxury businesses do not have when building models.
