Types of Machine Learning According to the usage of platform & method of learning.

Machine learning can be performed in different ways. It can be performed online or offline and it can learn by heart or learn from data. In this post, we will discuss Machine Learning types according to the learning method & according to the platform it uses.

According to the platform, a Machine learning algorithm learns it is classified into two categories.

1. 1. Batch Learning
   2. Online Learning

According to the learning method, an ML system can be classified into two categories.

1. 1. Instance-based learning
   2. Model-based learning.

 

Batch Learning:

This type of machine learning project we often see in different tutorials. In this type of system, the data is trained using a local machine or offline and this type of model or system is incapable to learn incrementally in the future. This means, if you build a model & launched it into production it will perform only the same types of data it has seen during the training period. If you need to update this model for new data then you need to create a new model with the new data & relaunched it to production. This model is built offline and this is why sometimes it is also called offline learning.

 

Online Learning:

This ML learning process is done online & this type of learning can learn incrementally. You can build this model by feeding one by one small part of your data. This learning method is continuous. This means it will learn automatically from new data while it is in production as it performs both learning & application online. So, there is no issue if you want to update the model for new data, it will learn automatically from the new data & don’t need to rebuild or relaunch again and again.

 

According to Learning Method

Instance-Based Learning:

Instance-based learning learns by or seeing the examples. For example, if you want to build a spam filter it will first need to have a number of spam emails that are already detected & filtered out. The model will go through every mail & find the most common words available in the mail. After doing that it will generalize all the spam words. Whenever a new mail arrives this model compares the similarity of words with the old generalized words & such it finds the mail is either spam or not.

 

Model-Based Learning:

This is the other learning we often see in tutorials. In this learning method, a machine learning model is created with the available data & launch in production. For say, you want to make a prediction of car price & you have some old data for this. Now, you can build a regression model with the old data by using scikit-learn or by using a basic regression equation. The algorithm learns through a model in this case a regression model. That’s why it is model-based learning.

conclusion

These are four types of learning methods you should need to know. Hope that you understand these & if you have any confusion related to this you can do comment or email me through the contact form.