# How to know what to seek: **Active Learning**

Today’s topic is

Active Learning.

Its a branch of Machine Learning to seek the right observations/teachers for effective learning. Now that’s quite obvious right! Even in real life we try to seek right teaching collaterals or a teacher, for effective learning.

So how this is similar or different in active learning. Lets define a Machine Learning use-case to understand the concept.

Suppose, you want to develop a model to predict likely loan default cases. In total, you have *10000* data points/rows of data and have only *1000 *labeled cases, out of this total data points.

As mentioned above, data is your teacher when you are developing a machine learning model. But for your defined use-case, you only have *1000* data points to teach your model. So what do we do with the remaining *9000* data points.

How do we select the right teachers from these 9000 data points?

*Active learning is the answer*!, which will help you choose the right data points/teacher to teach your ML model effectively.

The concept is to choose the cases with higher entropy (now, thats a technical jargon), so lets make it simple. Choose the cases with higher uncertainty (that’s much simpler, I think), or choose the cases with high error!

Next question is, how to find these uncertain data points?

Predict the unknown cases, with the ML model developed using labeled *1000* data points. Cases with largest error or highest uncertainty are the ones difficult to predict or learn and should be used to teach your model and should be labeled manually.

Why not to label all the cases, if even these uncertain cases are to be labeled manually?

Cost of labeling is **mostly** high which needs your SME’s time! So its very important that the SME’s time is used very effectively/efficiently, and the concept of active learning helps you to control this cost.

Now you can extend this concept, by choosing the right methodolgy to come up with the right teachers.

For example: Clustering could be used to avoid choosing similiar uncertain cases….and many more such tactics/ideas can be used.

“You are free to choose, but you are not free from the consequence of your choice, so choose wisely!”