Machine learning in schools #4

Part 1, part 2 and part 3 of the series can be found here.

Exams are over now and sports day is in a week later. Our protagonist Ryan belongs to the red house. The red house has a very good cricket team and has won the cricket tournament for the last 2 years. They’re determined to make it 3 wins in a row. However, this year their wicket keeper has a broken thumb and won’t be participating in the tournament. The house does not have another wicket keeper and training someone from scratch would be a huge risk.

Their coach decides to turn to Ryan. Ryan is among the best goalkeepers not just in the house but in the school as well. He has won many awards for the school’s soccer team. The coach believes that it would be easier to turn Ryan into a wicket keeper than to train someone else. This concept in machine learning is known as transfer learning.

In Transfer Learning, the knowledge of an already trained Machine Learning model is applied to a different but related problem. This is done by making use of similar features. Athleticism and ball catching are skills required in both sports and would come handy when switching from goalkeeping to wicket keeping. This reduces the training time of our model and gives better results. Another important reason why we use transfer learning is that for most real-world problems, we do not have enough labeled data points to train our model on.

If we think about it, transfer learning is really how humans function. We learn something and then we apply our knowledge to something else. If I can tell you what a rotten potato looks like I can also tell you what a rotten tomato would look like. If I can tell you liked this article, I can also tell that you will like the next one. Stay tuned.

~ happy learning