Types of machine learning for Data science:

 Types of machine learning for Data science:

What is machine learning?

As we all know the machines, or the computers work based on the instructions which we give. Unlike the humans' machines cannot think and act based on their past experiences.

Machine learning is a process in which sample data is given and the machine is trained. After that the machine takes decision or makes predictions based on the past data without programming.

Types of machine learning:


Types of Machine learning Algorithms

1) Supervised Learning

2) Unsupervised Learning

3) Semi-Supervised Learning

4) Reinforcement Learning

1) Supervised Learning

In Supervised Machine Learning the machine is trained using labeled training data. Both input and output are clearly specified with labeled data. It is similar to learning under the supervision of a teacher.

Some of the real-world applications are classifying spam mails from your inbox, fraud detection, image classification etc. There are two types of supervised learning. They are classification and regression.


Supervised Machine Learning

 The working of supervised machine learning can be easily understood by above diagram. First step is a dataset of different shapes are given. It is labeled as square if it has four equal sides, if it has three sides it is labeled as triangle and it has no sides it is marked as circle.

Then the model is trained using the labeled data and the test data is given for prediction. Finally, the model has to identify the shape.

2) Unsupervised Learning

In supervised machine learning the machine is trained using labeled data. But many a times we won’t be having labeled data. In that case unsupervised machine learning is used. 

So here the model has to discover the hidden patterns or group the data on its own without any human intervention. It is similar to the human brain learning new things.


Unsupervised learning

In the above diagram unlabeled data set of different shapes is given. Then the machine interprets the dataset and group the data according to the similarity between the images. Then the algorithm is applied and the data is divided based on the similarities or differences among them.

3) Semi-Supervised Learning

Semi-supervised learning is in between supervised and unsupervised machine learning. In supervised learning the machine is trained using labeled data, in unsupervised learning the unlabeled data is given and the machine has to group the data based on similarities and differences.

In supervised learning the data is labeled manually by the data engineers. The disadvantage of supervised learning is when the volume of the data is large it becomes very costly. The disadvantage of unsupervised machine learning is the output may not be accurate since the input is not labeled.

To overcome the disadvantages of supervised and unsupervised learning semi-supervised learning is introduced. Here the input is a combination of both labeled and unlabeled data.


Semi-supervised learning

The semi-supervised learning can be understood from the above example. Here the input data is unlabeled and partially labeled data. With the help of labeled data and grouping of data it makes the predictions. It is similar to a student learning with a little help of the teacher.

4) Reinforcement Learning


Reinforcement learning

In reinforcement learning an agent is placed in a environment. The agent has to take an action and has to get maximum rewards. Here the path is not given so the machine has to take decision based on experience and the rewards.

In the above example the dog is the agent, and it is put in the environment training. So, if it obeys the trainer and takes the action it gets the rewards and if it does not obey it doesn’t get the rewards. 

Based on the positive feedback and the negative feedback the dog has to make observations and act accordingly to maximize the rewards.

Post a Comment