top of page

5 Reasons why you shouldn't take Machine Learning course

The growing number of Twitter and LinkedIn influencers shows why you should start learning Machine Learning and how easy it is to get started.

While it's always nice to hear words of encouragement, I like to look at things from a different perspective. I don't want to belittle anyone or discourage anyone, I'm just stating my opinion.


When you look at what these Machine Learning experts (or should I call them influencers?) are, I wonder: why do so many people want to learn Machine Learning?

Perhaps the biggest reason comes from not knowing what a Machine Learning engineer does. Most of us are not working on artificial general intelligence or autonomous cars.



When the machine first appeared, many people thought it would spark a new industrial revolution. Fast forward to today and many people say it's nothing more than words.

Don't get me wrong. Machine learning is a useful tool, but it is much more than that. And it would be an exaggeration to say it's something of a Swiss Army Knife - I'd consider it more of a seaplane. In data science course you will need to know more about machine learning.




Is machine learning better?

Let me also start the conversation by clearly explaining all the reasons why machine learning is especially good in 2021. If you understand ML, career opportunities are endless, whether you are an employee, freelancer or business owner.


Numerous free and affordable online training courses.

If you have a STEM degree, learning machine learning is a breeze.


One day literacy will be a basic requirement for all skilled workers. If you want to work in technology, expect that you will need to at least know how ML works. We are  not saying machine learning is a bad idea. It was actually one of the smartest choices I ever made. What I'm about to share with you is the least desirable thing that no one tells you about machine learning.



Data related questions

As seen in the AI ​​Requirements section, machine learning relies on many other key factors. This framework covers everything from data collection, data storage, data transfer and data transformation. It is important that you have a strong track record of taking these first steps, otherwise it will be impossible to have reliable data.


Why is this important? You know the saying 'garbage in, garbage out': The effectiveness of your machine learning is limited by the quality of your data. Therefore, it is very important to have reliable data at the beginning. Not only do you require  your data to be reliable, you also need enough data to leverage the power of machine learning. If these two parameters are not controlled, you will not see the full power of ML.



Technical debt


Maintaining a machine learning environment is difficult and expensive. There are several types of “debts” that need to be considered, especially when developing machine learning methods:

Credible Debts: Credible debts refer to debts that come from uncertain sources and are based on raw data. Simply put, this refers to the cost of maintaining multiple versions of the same system, outdated features, and unused databases.


Debt Analysis: This refers to the idea that ML systems often influence their behavior as they are updated over time, resulting in indirect and implicit feedback.

Configuration Debt: The configuration of a machine learning system has the same debt as other systems. Creating small drawings should be easy, it should be difficult to make manual errors, and it should be easy to see the difference between different styles.

If you want to learn more about the technical debt behind machine learning systems, you can find the full article here.


Learning machine learning alone will not get you where you want.

ML is a very small part of the job if you have time:


if you want to give business advice; If you want to review and market the product  and call and if you want to use the information to help you create better products and services.


This is essentially a re-emphasis or expansion of previous posts. If you are studying machine learning, I think you will want to find a job in data science. You think you will become a data scientist. You think you'll like it because it's the best job of the 21st century. But one of these three things could be wrong.

So someone could learn mechanics and find themselves working 60 to 80 hours a week at a hedge fund. And you know, it's high speed and you can't really mess it up and you get paid a lot of money.

But people are not happy and it is not good to see people doing this. And they say, wow, this money can't bring you happiness if you do that. When you think about response skills, you have to figure this out for yourself. It's like understanding who you are as a person. 


What impact do I want to make in my life? So what kind of life do I want to live and how do I want to change the world?


Now search for features and rules that work like a car. Use it as a way to overcome those things and you will have those things. Once you know what it is, figure out what you need and the skills to get there.


Most likely if it is in the information field, machine learning will be a part of it.


Better Alternative


Finally, machine learning should not be used when simpler methods are available. In my previous article, “If You Want to Become a Data Scientist, Do Not Start with Machine Learning”, I emphasized that machine learning is not the solution to every problem.


A simple solution that takes 1 week to build with 90% accuracy will always be preferable to a machine learning solution that takes 3 months to build with 95% accuracy.


Well, you should start with a simple solution that you can implement and that will clearly show whether the benefits of the next step outweigh the costs.


If you can solve your problem using a Python script or SQL query, do that first. If you can solve your problem with a decision tree, do that first. If you can solve your problem with a backup line, do that first. 


 Interpretability


Descriptive interpretation focuses on understanding the relationships between data variables. Essentially combining machine learning and neural networks,

Machine Learning is a predictive model. It is excellent at making predictions and far exceeds the predictive power of traditional models such as linear/logistic regression.


However, these models are a black box when it comes to understanding the relationship between structural change and dynamic change. While you can understand the basic mechanisms behind these models, it is still unclear how they reach their final conclusions.


While some techniques, such as principal factors and correlation matrices, are useful, they still do little to understand the relationships in your data. Machine learning and deep learning in general are good at prediction but lacking in details.



Machine learning sounds difficult


Many internet users say: Getting started with machine learning is very easy. Just download the Titanic Dataset, copy the 10 lines of Python code from the tutorial, and start Machine Learning.

But it is difficult  to imagine that anyone would pay for such information. That's why you need to dig deeper.


And the deeper the level, the harder it is. It's very important to have a good mentor so you don't have to figure everything out on your own. Interning is also a great way to grow as an engineer.

I wish someone had told me this at the beginning of my career. I had to spend countless hours trying to keep up with colleagues working in other areas of computer science.


0 views0 comments

Comments


bottom of page