4 Mins Read

Machine Learning

Machine Learning

What is Machine Learning?

Machine learning is a sub-branch of artificial intelligence that studies algorithms behind computer programs capable of learning  automatically from experience. Machine learning models build their logic based on the example data they receive and rely on algorithms to improve their predictions based on their current performance; as a result, they can learn and act without explicitly programmed instructions.

Machine learning: what type of learning does it use?

Several learning methods are common  in machine learning. The algorithms to train models fall into one of the following categories.

  • Supervised Learning
  • Unsupervised Learning
  • Reinforcement Learning
  • Semi_Supervised Learning


It is better to use an example to explain each category. Let’s consider the experience of a young child exploring his world for the first time. A child’s learning process is similar to that of a machine. Let’s say we want to teach our child the difference between an orange and an apple. In supervised learning, inputs and outputs are clearly defined, and an observer provides data and information (labels) to the learner. So we show each fruit to the child and repeat its name. Unlike supervised learning, which uses specific data, unsupervised learning does not require labels, and the learner must look at a particular pattern in the data to learn categories. So in an unsupervised learning example, we put a number of apples and oranges in front of the child without him knowing their difference beforehand and ask him to sort them into two separate categories and monitor his process. Reinforcement learning is based on rewards and punishments. In this learning method when an action is performed correctly, the model gets a reward and gets punished for incorrect actions with a reduced score. In this type of learning, the model attempts to increase its scores, which indicates that the model is learning correctly. In the case of a child, the parent shows the child an apple and asks him to identify it. A reward should be given to the child if he correctly identifies the fruit, and a warning if he incorrectly identifies it. As this process continues, the child becomes more adept at identifying fruits. Semi-supervised learning is a combination of supervised and unsupervised learning methods.

Why machine learning?

Recently, machine learning has gained a great deal of attention. Despite the discontinuation of some old methods, machine learning has demonstrated success in a wide range of fields. We are now able to easily solve many old and new unsolved problems. We have listed some of the several areas where machine learning has proved useful:

  • Finance, accounting, credit scoring, and algorithmic trading
  • Image processing, computer vision, face recognition, motion recognition, and object recognition
  • Computational biology, cancer tumor detection, DNA sequencing, drug discovery, and even drug cases
  • Energy production, market price forecast
  • Car manufacturing, aerospace

Natural language processing, voice, and language recognition programs.


Machine learning processes

1- Data Gathering

The first step in building a machine learning model is to collect data for training. To have a model that can learn all future problems successfully, the selected training data should be representative of all the possible combinations of outcomes. In machine learning, the dataset is usually divided into three parts: the main part is used for learning, and the validation and test datasets are used to evaluate the model.

2- Model selection

The second stage is the selection and training of a model. Various machine learning algorithms and models have already been developed and modified for improvements. Depending on the requirements of our problem, we can choose a model and train it.

3- Evaluation of a Model

Machine learning models learn various patterns and features of the provided data during training. Models can learn to perform different tasks, such as classification and regression analysis. After we had trained one, we then need to evaluate it using the test and validation datasets to ensure it has found the right patterns in the data.

4- Hyperparameter Tuning

Hyperparameters are parameters the model cannot estimate by itself and must be specified by the user before the algorithm runs. They are very important for the learning process, so choosing the proper value will help improve the learning of the model.


Related articles
History of AI
In the second part of a series of articles about the history of artificial intelligence, we look at important...
computer vision datasets
This article reviews famous datasets in the field of computer vision. ...
Dataset-Development-Lifecycle copy
Google has come up with a framework for data collection inspired by software development concepts in a 5-step cyclical...
From NLP to Computer Vision
In this article, we discuss the Attention mechanism and trace its history of use from natural language processing to...
TensorRT is a library developed by NVIDIA for faster inference on NVIDIA graphics processing units (GPUs). It can improve...
Generative Adversarial Network
A GAN (Generative Adversarial Network) is a framework to estimate generative models. This framework includes two models: generators and...
Subscribe to our newsletter and get the latest practical content.

You can enter your email address and subscribe to our newsletter and get the latest practical content. You can enter your email address and subscribe to our newsletter.