If you enjoy my content and want to get more in-depth knowledge regarding data or just daily life as a Data Scientist, please consider subscribing to my newsletter here.
As a Data Scientist, machine learning is our main tool to solve business problems and one reason we are employed within a company. However, would it enough to only use machine learning without any math knowledge behind machine learning algorithms? In my opinion, you need to learn the math behind machine learning.
And here is some arguments to support my claims:
- Math helps you select the correct machine learning algorithm. Understanding math gives you insight into how the model works, including choosing the right model parameter and the validation strategies.
- Estimating how confident we are with the model result by producing the right confidence interval and uncertainty measurements needs an understanding of math.
- The right model would consider many aspects such as metrics, training time, model complexity, number of parameters, and number of features which need math to understand all of these aspects.
- You could develop a customized model that fits your own problem by knowing the machine learning model’s math.
Even though we understand how important math is, the main problem when learning math is how high the learning curve. In my experience, many people give up learning math because they fell into pitfall learning that hinders their development.
In this article, I want to outline what mistakes you should avoid when learning math in machine learning. Let’s get into it.
1. Did not know what math topic is necessary for machine learning
It is admirable already to start learning math because the beginning is always the hardest. Although the intention is there, the problem is identifying which math topic to know when learning math for machine learning. Math is a broad topic, after all.
The mistake I often encounter is that people start to learn the math topic that did not touch the machine learning requirements and did not research enough what math topic support the machine learning field. I have an experience where I ask someone how you start studying math for machine learning – and their answer is by opening their high school math material; clearly, a wrong place to start learning.
My suggestion is to start learning this basic math topic for machine learning:
- Linear Algebra
- Multivariate Calculus
- Optimization Methods
For supporting your study process in these math topics, I would share the link to the additional resources at the end of this article.
2. Did not ask for help
The responsibility of learning is lying on yourself, but it is always fine to asking help from others. It was not a shameful thing when you did not know something, especially about math. You see all the memes out there; math is the personification of the hardest thing you would ever meet.
I have been in the position of not understanding the math concept presented in the book. I have tried to search all the material, paper, and book, but it just never clicked to me somehow. in the end, I decided to ask for help from someone. Endlessly scourging for the material is just a waste of time, after all. My acquaintance explains the machine learning math concept is way better than any material I ever read, I understand perfectly what he explained, and until now, it is still engraved in my mind.
I really encourage everyone to ask for help if they did not understand, especially those who start their journey in the Data Science field and Machine Learning math. You could start asking questions with the people you look up to on social media, such as LinkedIn or YouTube. Stackexchange or Reddit also a great place to start a mathematical discussion, although it depends if people would answer your question. Nevertheless, try to ask for help if you did not understand something.
3. Jumping to learning Machine Learning Math without understanding the Machine Learning algorithm concept
You already know what Math topic to learn, but it is still a broad thing to learn. Remember, we want to learn about math for machine learning, and not just any math topic; that is why we need to relate it with the machine learning algorithm.
This is a mistake that I once made in my early times. I know that I need to understand math to become a great data scientist, so I learn about Linear Algebra. However, what I learn did not translate to understanding machine learning math because I cannot relate linear algebra math with machine learning math. In this case, I try to change my approach by understanding the machine learning concept as my starting point.
For example, in my student time, I learn machine learning coding by importing the Linear Regression model. I know how to use the model, but I did not exactly understand how it works. To understand the Linear Regression concept, I start looking for the learning material, and from this, I am introduced to many new terms, such as Linear Function. When I started to understand the Linear Regression concept, I try to delve deeper by learning the math concept in each new term I found out. With this approach, I capable of understanding math better.
4. Focusing on Math for Data Science instead of Math for Machine Learning
While Data Science and Machine Learning is an intertwined topic, they inherently have different math concepts that support them. The fundamental mistake is learning a math concept that focuses on Data Science instead of Machine Learning.
What is the difference between Math for Data Science and Math for Machine Learning? It is the purpose. When we learn Data Science, this field analyzes the data we have and tests the hypothesis to validate our assumption. This is why we are often learning about probability and Statistics when we are learning Data Science because we rely on probabilistic math to conduct the hypothesis testing.
However, math in machine learning is different. They focus more on Linear Algebra as a basic process for many models we used and the Multivariate Calculus for numerical optimization, which become the backbone for almost the machine learning algorithm we used. For example, Logistic Regression is based on the Linear function (hence Linear Algebra). The coefficient is optimized via Maximum Likelihood Estimation (hence the need for Multivariate Calculus).
I would not say it is a fatal mistake to focus on Data Science math because it is still useful in your everyday data activities. Moreover, in my opinion, Data Science math is a prerequisite you need to know before learning more about math for machine learning.
5. Stuck in the "School-Days" Way of learning
Humans are creatures of habit, so we love to do things we are most familiar with. This includes our way of learning, where we are taught to learn by using only the pen and book – which means we are only focused on the theory and answering textbook questions. There is nothing wrong with this learning way if you plan to specialize in machine learning academia or research. Still, in industrial cases, you would need a different approach.
In the business environment, data scientists need to have a fast-paced, flexible, and applicable mindset. Learning math for machine learning would be similar; you need more concerned about the intuition and application behind the math instead of the theorem. Current Technology has improved so much that all the laborious work of manually working through the problems is not essential. It is way more sense to rely on computational power instead of writing every equation on paper.
Additionally, you could use computational libraries such as NumPy to supporting your learning process. This package is developed to make your life easier so that all the equation you need is already inside this one package.
Math for Machine Learning Resources for Learning
If there is a mistake, there is surely a correct way. If you have learned the mistake to avoid when learning math for machine learning, there are additional resources I want to share with you all.
- Mathematics for Machine Learning by Marc Peter (2020)
- Mathematics for Machine Learning by Garret Thomas (2018)
- Linear Algebra by Jim Heffereon (2020)
- Multivariate Calculus by Shurman and College
- Calculus for Machine Learning YouTube Video
- Optimization Learning YouTube Video
- Basic Mathematics Notation for Machine Learning by Jason Brownlee
Conclusion
Learning math for machine learning is important for many reasons, although there are some study pitfalls you could encounter; these mistakes are:
- I did not know what math topic is necessary for machine learning
- Did not ask for help
- Jumping to learning Machine Learning Math without understanding the Machine Learning algorithm concept
- Focusing on Math for Data Science instead of Math for Machine Learning
- Stuck in the "School-Days" Way of learning
I hope it helps!
Visit me on my LinkedIn or Twitter
If you are not subscribed as a Medium Member, please consider subscribing through my referral.