The Different Types Of Machine Learning Problems

Types Of Machine Learning

There are 3 main types of machine learning problems:

• Supervised Machine Learning
• Unsupervised Machine Learning
• Reinforcement Learning

Supervised Machine Learning

In supervised machine learning, there is usually labeled data that is used to predict a target attribute.

The main classes of supervised machine learning:

• Classification
• Regression

Classification is used when you want to take data on a problem and gain a general representation of that data to predict a label.

For example, in the popular Iris dataset, we are given the following information:

1. sepal length in cm
2. sepal width in cm
3. petal length in cm
4. petal width in cm
5. class:
— Iris Setosa
— Iris Versicolour
— Iris Virginica

In this scenario, the class of flowers would be our target attribute.

So basically what we are doing is… we will try to take previously collected data on the flowers and try to predict future samples of irises based on their petal and sepal sizes.

This is one example of a classification problem. Other examples would include:

• Titanic dataset to predict which passengers would survive
• SMS dataset to predict spam in text messages

Classification Algorithms:

• K-Nearest Neighbors
• Logistic Regression
• Decision Trees
• Random Forest
• Support Vector Machines
• Neual Networks
• Gradient Boosting (xgboost, lightgbm)

Regression is used when you have continuous data and would like to predict a data point along that continuity.

Regression Algorithms:

• Linear Regression
• K-Nearest Neighbors Regression
• Random Forest Regression
• Gradient Boosting Regression

Another type of supervised learning is called semi-supervised machine learning. Semi-supervised learning uses a combination of supervised and unsupervised methods when there is only a portion of labeled data but the data have similarities that can be grouped.

Unsupervised Machine Learning

Unsupervised machine learning differs from supervised machine learning in that there are no tags for our data.

Since there are no labels, the algorithms usually find clusters of similar data based on their attributes.

Unsupervised Algorithms:

• K-means clustering
• Principal Component Analysis

Reinforcement Learning

Reinforcement learning is a type of machine learning algorithm that uses an agent that acts based on a policy gradient.

Examples of reinforcement learning would include game playing agents that learn to beat a level over time by watching the pixel values.

Reinforcement Learning Algorithms:

• Q learning

Conclusion

This post was to give a general representation of the different areas of machine learning. I hope it wet your appetite for machine learning and found it interesting.