Types of Machine Learning

What is ML?

Machine learning is the process of automating and improving the learning process of computer systems based on the experiences gained on the DATA(Input) we feed + Output without being actually programmed by allowing the machine to develop its own logic during training.

Types Of Machine Learning:

the following are the types of Machine Learning Algorithms:-

1. Supervised learning.
2. Unsupervised learning.
3. Reinforced learning.

1. Supervised Learning:

In Supervised learning, the machine is trained on labeled data. The Machine learning algorithm is provide with a training dataset to train/learn. This training dataset is a subset of the bigger dataset and thus gives the machine an opportunity to learn and train itself. As the training dataset is a subset of the bigger dataset, hence it has very similar parameters to the bigger dataset.

The algorithm then finds relationships between the attributes and established cause and effect relationships between the variables in the dataset. so, After sufficient training, the algorithm has an idea of how the variables of the data are related to each other and what can be expected after changing some parameters. This logic is then deployed for use with the final bigger dataset.

The supervised learning algorithm thus keeps on learning and improving itself by changing the strength and weights of the attributes continuously. Supervised learning is commonly used in marketing and sales, sentiment analysis, customer lifetime value, recommendations, time-series market forecasting, human resource allocation, fraud detection.

Some common Supervised learning algorithms:

1. Linear Regression
2. Nearest Neighbor
3. Gaussian Naïve Bayes
4. Decision Trees
5. Support Vector Machine (SVM)
6. Random Forest

2. Unsupervised Learning:

Unsupervised machine learning also has the advantage of being able to work and learn with no previous learnings and prejudices. This means that human intervention is not require to make the machine learn. This however allows the user to work on much larger datasets. The relationships between data points are recognize by the algorithm in an abstract manner.

The auto-generation of internal structures and connections without human intervention on the basis of the dataset itself is what makes unsupervised learning algorithms very robust. Unsupervised learning algorithms can thus adapt and dynamically learn about hidden patterns in the data which cannot be identified easily. Some of the industrial applications of unsupervised learning are visualization, finding association rules, clustering, anomaly detection, and dimensionality reduction. 

Example of clustering on Iris Dataset:

Some common Unsupervised learning algorithms:

1. Hierarchical clustering
2. K-means clustering
3. K-NN (k nearest neighbors)
4. Principal Component Analysis
5. Agglomerative
6. Probabilistic

3. Reinforced learning:

Reinforced learning is all about taking suitable actions and getting rewards for appropriate results and learning from it. It’s like getting a chocolate every time you give a correct answer. This reward system trains the model to act in a way to generate desirable results as per the training. Supposing the aim of the robot is to get the reward, that is the chocolate by avoiding maximum hurdles.

The robot learns by trying all the possible paths and then choosing the path which gives it maximum reward with the least hurdles. In Reinforced learning, each right step will give the robot a positive reward and each wrong step will give a negative reward. The sum total reward will be calculated when it reaches the destination and gets the chocolate.

Broadly there are two types of reinforced learning methods:

1. Positive Reinforcement: It is the event of positive addition in the total reward, occurring due to a particular behavior which in turn increases the strength and the frequency of the behavior. It assists the model to learn the correct path and act as per the learning in the future. On the other hand, too much Reinforcement can lead to a decrease in the efficiency of the model.
2. Negative Reinforcement: It is defined as the negative addition in the total reward, occurring due to a certain behavior which in turn decreases the strength and the frequency of the behavior. It assists the model to understand and learn which path to avoid and act accordingly. Although it only provides enough learning to encounter the present issue with a minimalist approach. 

Reinforced learning is predominantly used in robotics for industrial automation, machine learning, and data processing, systems that provide custom instruction and materials according to the requirement.

Written By: Jeet Barot

Reviewed By: Krishna Heroor

If you are Interested In Machine Learning You Can Check Machine Learning Internship Program
Also Check Other Technical And Non Technical Internship Programs