There are different algorithms in Machine used for different types of datasets. To solve a specific problem it is important to identify the problem and choose a suitable mode of learning of the problem.
To select the type of machine learning we must understand the data types by analyzing them. We have already discussed the process to start a machine learning project. So now lets see the types of machine learning.
Machine Learning types are classified into three types:
- Supervised Learning
- Unsupervised Learning
- Reinforced Learning
Supervised Learning is mostly used in practice to train the model ad this type of learning are used for the data with labels. It is very easy to understand and can be implemented by using simple mathematical functions like linear regression.
This module fits itself with the data and more data results higher accuracy but we need to consider overfitting problem.
The main categories of Supervised learning are Classification and Regression.
Supervised Learning is implemented to classify the diseases like pneumonia, cancer, fractures of bones and many more by simply providing the classified data. Spam Classification, Weather Forecasting, Image Classification and business prediction can also be done using Supervised Learning.
Unsupervised Learning is used to train a model using the information that is unlabeled. The algorithm will find the similarity in the data and classify it into a cluster. Clustering Technique in Unsupervised learning is mostly used in practice where the algorithm will divide the data in the specific number of clusters.
Dimensionality reduction is also used in unsupervised learning where high-dimensional images are transformed into low-dimensional so the features can be extract easily.
Unsupervised learning is used mostly for large data like customer segmentation of Stores, User Classification in Social Media and is highly used by business analyst. Companies like Facebook, Amazon uses mostly this type of learning.
This is a type of learning where the machine learns form the mistakes. Here data is not externally fed by the user but data is generated by the machine. The machine gets to know what is right and what is wrong and improves itself over time.
A reward system is considered in this system. Reward is given to the machine if it perform correctly.
We can simply look an example of training a dog where the dog is rewarded if the task given is completed and not rewarded if the task is not completed. Students of Stanford University used this model to train a helicopter. The helicopter crashes many times and at the last learned to fly itself.
The figure is an example of a AI game module where a reward is given if the system makes the correct move. The no of episode simply tell us that how long iterations it is require to train the module itself.
After getting the basic knowledge of machine learning come and learn about Data Preprocessing to start your Machine Learning career.