How a Single Mistake Wasted 3 Years of My Data Science Journey

By Pranjal Saxena, Data Scientist, Top Writer in Artificial Intelligence

Adult and child working on a computer
Photo by Julia M Cameron from Pexels


I started my data science journey back in 2016. At that time, machine learning boot camps were much popular. The only requirement of the course was basic Python knowledge. I felt so happy because I used to play with Python codes at that time, so I opted for the course.

The usual practice of human beings — when we’re learning something new — is to try and collect as much material from different places. We may think that these resources will help us learn more. And, the more variety of materials you have, the more choice you can have to select the right resource. But, this is wrong.

Most of the resources I collected back in 2016 are still in my D: drive, and I haven’t watched them yet. But, yes, there is one course that I covered 70% of, which was the machine learning boot camp course. And the reason I finished that course was — the course was easy to follow and everything was simple and straightforward.

When we are in the machine learning area, we may have the fantasy to build a model without knowing the ideas behind the algorithms. And that Bootcamp course primarily focused on the implementation part and not on the concept part, and that’s where I made the big mistake. I was in the false hope of learning machine learning.

Machine learning is not about importing libraries, feeding data, and getting results. It’s more of the knowledge behind the algorithms that helps us select the correct algorithm for our data to get a better outcome.


Self-Paced Courses Are Just Sleeping Pills

The self-paced courses provide us with passive learning, which is not suitable for machine learning. Here, you can find a neural network with different layers, activation functions, back prop, forward pass, and many more things that a single person can’t cover in that self-paced course. And nobody will share those tricks in the self-paced course — usually, those are free or much cheaper.

I was like everyone else, trusting the self-paced courses of machine learning where everything is done by another person. Here I am sitting, just watching the video tutorial to build a complete machine learning model without trying even a single line of code myself. Then, getting the false hope of learning machine learning by just covering those videos.

I am not against self-paced courses. It depends on the area of learning. A self-paced course for learning guitar might be a good one to practice because we only need a guitar in hand to try it out.


Industry Experts Are the Right Choice

After three years of trying self-paced courses from different resources, I found that I cannot even decide the best suitable algorithms for specific problems. I wasn’t aware of the simple modeling process. All I was aware of was importing certain libraries and passing data to get the output of regression and classification problems. But, how to interpret those outputs, I had no clue.

Finally, one day, one of my colleagues suggested joining an institute where industry experts would teach us. That day I was like, it is just a waste of money because they were charging ten times more than those self-paced courses.

Anyhow, I joined the online live classes and observed the change. That active learning through live classes gave me a boost. The step-by-step process of building a neural network gave me another boost. And the main benefit was the daily assignment given to us.

The main differences I felt were the following: I could get in touch with the instructor to clear up any doubts; work on real-time industry scenarios, enjoy professional-quality content, and I was required, which we can’t get from a self-paced course.

Most of the freely available courses on the internet are not well structured, and that’s the secret of machine learning. There is a step-by-step process to build any model. First, we have to learn about the algorithms, get the data, clean the data, find the best features, split the data, train the model, validate the model, and then do inferencing. We can’t miss any single step if we want better results.


Final Thoughts

In the above article, we discussed the importance of learning from an industry expert to know the proper flow of machine learning modeling. Otherwise, you will keep trying many free resources without understanding the adequate flow for modeling — so, they are of no use.

I hope you liked the article.

Thanks for the reading!

If you want more such exciting articles on data science and technology, here’s my newsletter.

Thanks to Anupam Chugh.

Found this story interesting? Follow me (Pranjal) on Medium. If you want to reach out to me with private questions do connect me on Linkedin. And, If you want to get more exciting articles on data science and technology directly to your mail then here is my free newsletter: Pranjal’s Newsletter.

Original. Reposted with permission.


Source link

Leave a Reply

Your email address will not be published. Required fields are marked *