# A Quick Overview Of Machine Learning Tasks🚀

In this article, we will learn an overview of machine learning techniques and how they perform well based upon our problems, including simple description, visualization, and example for each one.

Here, the techniques of what we are going to cover.
• Regression
• Classification
• Clustering
• Dimensionality reduction
• Reinforcement learning (Ranking)
Those are all the foundations of machine learning, where you will get some essential good skills and ideas about Machine learning tasks to help begin your  machine learning carreer.

## Regression

The regression task comes from Supervised machine learning. Which can help us to predict (expect continues values) and explains objects based on a given set of numerical and categorical data. For example, we can predict house prices based on house attributes such as several rooms, size, and location.

In mathematical terms, the regression method provides us a linear line belongs to the equation of Y = aX+b to model a dataset. Here we are taking the X (Dependent variable) and Y (Independent variable) data points for train our linear regression model. The best observation line can be found by calculating the slope (m) and y-intercept (b) values.
In the above, an image represented a linear regression task, which contains two different data points considered as a person's height and weight.
If the height is measured on the X-axis and the weight measured on the Y-axis.
The data points from the two features have been plotted in blue.
By using formula Y=aX+b, we can draw a linear line represented with a red color that is known as a predicted line.

Here, the regression analysis performs with various effective algorithms,
• Simple linear regression
• Multiple linear regression
• Polynomial linear regression
• Decision tree
• Support Vector machine (SVM)
Applications
• Risk assessment
• Score prediction
• Market forecasting
• Whether forecasting
• Housing and product price prediction
• Digital personal Assistance (Siri and YouTube)
• For better understanding, see below an image.

## Classification

Classification is a supervised learning task used when we need a limited set of outcomes. It generally provides predicted output values, which may be  True or False.  It performs in the way of two types such as Binomial and Multi-Class. For example, we can find whether a received email is Spam or Not.

Here, we have data points belonging to three different fruits like blueberry, lime, and apple. We can classify them by using machine learning classification techniques. The three fruits are represented by three various related colors such as blue, orange, and yellow respectively. Then, we draw a linear line between them for clear separation and see how they are presented.

Classification supported Algorithms
• Logistic regression
• KNN (k-nearest Neighbor)
• Naive-Base
• Discriminant Analysis
• SVM (support vector machine)
• Decision tree
• Neural network
Applications
• Image classification
• Email spam detection
• Fraudulent detection

## Clustering

An unsupervised learning technique performs as a way of grouping related items together into various clusters. It doesn’t have any output information for the training process. Nonetheless, the clustering algorithm will define the output. Here, the best clustering algorithm is K-means, where the K represents the number of clusters.
Here, we have three different types of data points represented by three colors. Clusters can organize a bunch of data based on their characteristics.

If above mentioned the three different data points separated into three different clusters and named cluster 1, cluster 2, and cluster 3

Algorithms
• Singular-value Decomposition (SVD)
• Hidden Markov model
• K-means
• Gaussian Mixture
• Neural networks
Applications
• Recommender system
• City planning
• Targeted marketing
• Customer segmentation

## Dimensionality Reduction

An unsupervised learning task used to reduce redundant information from large data sets. It provides less computation and reduces training time.

Then it can filter low variance and high correlations of them. When we have a large dimension, some algorithms do not work well.
Here, the popular algorithm PCA can help us to reduce the dimension of the future space.

Let’s have a simple example for dimensionality reduction tasks.
Here, we consider above the multidimensional cube, which includes lots of different data points itself.

So, we can do slicing operations to separate dimensions for our specific task represent as dimensionality reduction.
In task 1, we separated the front side of the cube that has been represented 2 Dimension view contains 100 positions.
As same as we also separate 50 positions only as in the 1-dimensional view from the top of the cube.

Algorithms
• Principal component analysis (PCA)
• Linear Discriminant Analysis (LDA)
• Generalized Discriminant Analysis (GDA)
Applications
• Text mining
• Face recognition
• Bigdata visualization
• Structure discovery
• Image recognition

## Reinforcement Learning (Ranking)

A reinforcement learning technique doesn’t have target variables, instead it's a reward-based system. Unlike Supervised learning which has the training data, it includes an answer key, the reinforcement agent will determine what to do to perform the given task.

Here, we consider a toy example: navigating to the exact point in the 4 * 4 matrix.
The car starts from the lower right corner of the maze. There are four action choices (right, left, top, and town) to reach an exact location point. The matrix diagram includes  firewalls and non-fire walls. If the car chooses the wrong action, then it gets fired, otherwise it moves the next matrix. Those results are calculated as a negative and positive reward respectively.

The first action is either you move a top or left matrix. If you move the left side then you will be getting fired and it will receive reward -1 to the car, Otherwise reward 0 to move next. Now, the car can determine which side is good for it to move further. The car is taking multiple Iterations until reaching an exact location based on its experiences. In iteration 5, the car finally found the correct location represents in green color.

Reinforcement Learning Algorithms
• Q – Learning
• State – Action -Reward- State- Action (SARSA)
• Deep Q Network (DQN)
• Deep Deterministic Policy Gradient (DDPG)
Applications
• RL can be used in space exploration and Navigation (Rover)
• RL can be used in Motor control
• RL can be used in Sequence learning for visual captioning
• RL can be used for better Decision-making tasks.

## Summary

In this article, we studied Machine Learning techniques, several algorithms available, and applications based upon ML tasks.  