Learn but Do Not Overfit!

Information Overload — When there’s too much to learn in Machine Learning

A beginner in Machine Learning is just like one’s model. With access to courses on Coursera, tutorials on YouTube, competitions on Kaggle, and GPU on Google Colab, the newcomer is just like one, ML model, having access to MNIST dataset, TensorFlow and other python libraries.

Just like the model iterates through the whole dataset, the newcomer enrolls in a course, watches a few tutorials, and reads some articles. While advancing through the dataset, the model then adjusts its weight to improve its accuracy, essentially like the beginner sticking to the best learning method, be it a course or some project.

Learning in ML is gradual and uncertain, for both the beginner and their model (unless the machine has Nvidia’s Titan V GPU, and the person has extraordinary intelligence).

Just like a newcomer makes sure their model does not overfit, one should also care for their mind. Burdening oneself with courses and projects will not make one an ML specialist, just like training the model on too much data would not improve its accuracy.

With rapid growth in Machine Learning and increased competition for internships and jobs, one feels intimidated by the knowledge and skillset he/she must acquire to be an intermediate in the field. Aiming to flourish in Machine Learning, one burdens him/herself with courses, projects, and goals more than their potential.

The fear of not being able to compete and grow among fellows causes work-anxiety among students.

This stress and anxiety not only reduce productivity but is also detrimental to one’s mental and physical health.

Instead, like an algorithm’s advancements and dataset’s evaluations improves the model over time, someone aiming for success in this field should learn slowly but effectively. Maybe start with taking a course and later sign up for some related project.

Summarizing the discussion, one may conclude that evaluating one’s learning regularly, appropriate workload, and of course, regular breaks from work can be one’s key to success, because even Google Colab has a timeout!