MIST101-4: Overfitting, Underfitting and Regularizations

2020/2021: 1 | 2 | 3 | 4 | 5


Image of MIST101-4

From the previous workshops, we now understand how a machine can learn from examples (data). But like humans, can machines refuse to learn or maybe fail to learn the right things? The answer to this question is one of the day-to-day struggles in an ML engineer’s life! So in this workshop, we will dive deeper into the concepts of over-fitting, under-fitting, and regularizations. Understanding these concepts will enable us to answer the question I asked before and explore the practical difficulties faced when training an ML model!

This workshop also includes a coding tutorial that will teach you practical skills such as identifying problems from analyzing training curves, cleverly adjusting parameters to improve model performance, etc. You will also learn to implement GPU training on Colab, to efficiently train your models.

Good News! We have made the process of participating and winning prizes a LOT simpler. After watching the workshop videos, you can test how well you have understood the concepts by participating in a short online quiz which consists of MCQs and short answer questions. This will not just help you reinforce what you have learned but also help you win prizes!

The quiz can be found here!

As always, we are partnering with Amazon Prime Student to provide gift cards worth 25 CAD each when you answer them correctly* ! So stay tuned for our upcoming workshop which will be released on our YouTube channel on October 31st, 2020.

Wait a minute, what time is it?…. We value the fact that you are busy, and hence we’ll upload the recorded videos on our YouTube channel, so that you can watch it when you have time!

Presentation slides are here and the Colab notebook can be found here.

Event Details

Topics to be covered:

  1. Recap: gradient descent, backpropagation, and model training
  2. Overfitting
  3. Underfitting
  4. Methods to keep the model ‘fit’
  5. Regularization in linear models
  6. Coding tutorial

How to win prizes:

  • Participate in the quiz.
  • Answer the MCQs and short answer questions to the best of your knowledge.
  • Hit submit.
  • That’s it!

The quiz should take no longer than 10 minutes, provided that you have watched the videos!

We’ll contact you if you win!

References:

Mishra, M. (2018, June 10). REGULARIZATION: An important concept in Machine Learning. Retrieved October 27, 2020, from https://towardsdatascience.com/regularization-an-important-concept-in-machine-learning-5891628907ea

Overfitting in Machine Learning: What It Is and How to Prevent It. (2020, May 23). Retrieved October 27, 2020, from https://elitedatascience.com/overfitting-in-machine-learning

* An Amazon gift card worth 25$ will be provided to the top submissions.