Introduction to Machine Learning, the Easy way
In this blog, we will start with the basics of Machine Learning, terminology related to Machine Learning, and some interesting facts about it.
Introduction
Machine learning is the study of computer algorithms that allow computer programs to automatically improve through experience ~ By Tom M. Mitchell. It is a branch of Artificial Intelligence and Computer Science that relies heavily on data to improve its learning accuracy.
Just like a student who is trying to learn new topics, after learning from the information, the student gives a test, gets its score, tries to evaluate its performance, and makes sure not to repeat the mistakes, similarly, the machine tries to learn from experience and trains itself. In short, it’s all about learning and improving from past experiences, to be more specific machine needs to generalize its learning rather than cramming the data.
How old is Machine Learning?
Although Machine Learning was known in the 1950s, it became so popular nowadays, due to the arrival of super cheap computing and memory capabilities, that happened with the arrival of public clouds offerings, as well as the availability of improved hardware — powerful laptops and desktops. As we said earlier that these ML models learn from data, with huge advancements in the areas of Big Data, it became quite easier to handle the data as well & apply analytics to it.
Some examples of Machine Learnings:
What you think today becomes a reality tomorrow! So true in the case of Machine Learning right? From autonomous cars to our smartphone assistants Google, Siri, Bixby. From categorizing our mails to detecting spams, from giving us product recommendations, from image recognition to language translation, from fraud detection in our financial systems, and innovations in the healthcare industry — machine learning has played an important key role.
Types of Machine Learning:
Machine learning is classified into three categories :
- Supervised Learning
- Unsupervised Learning
- Reinforcement Learning
Supervised Learning:
When a machine tries to learn from a dataset that contains inputs as well as the desired outputs, this kind of learning is known as Supervised Learning. Generally desired output acts as a supervisor that helps the machine to learn from a labeled set of data.
It is further classified into two categories:
1. Classification
2. Regression
Classification:
When a machine tries to classify the data into different discrete categories, it is known as Classification. Some examples:
- Predicting whether it will rain or not, based on the weather dataset.
- Predicting person’s choice of meal — Vegetarian, Non- Vegetarian, Vegan.
Regression:
When a machine tries to predict a continuous value rather than a discrete class, it’s known as Regression. Some examples:
- Predicting the amount of rain during the rainfall, based on the dataset.
- Predicting the house prices, based on the dataset.
Unsupervised Learning:
When a machine tries to identify patterns in the dataset on its own without any supervision or labeled dataset.
Examples are:
- Clustering
- Association between data
Clustering:
Analyzing the dataset and find natural groups or clusters in feature space. For example, Customer segmentation.
Association :
Machine tries to find out the relationship between the data points. For example:
Market Basket Analysis — Using association rule, to find out which items are bought together.
Reinforcement Learning:
This kind of learning enables an agent to learn in an interactive environment where it tries to maximize the reward and minimizes the punishment by learning from the feedback from the trial and error approach experiments. This method assigns positive values to the desired actions to encourage the agent and negative values to undesired behaviors.
Its major applications are in Gaming and Robotics.
Challenges while using Machine Learning:
Till now in this blog, we have gone through an introduction to Machine Learning, but let’s now briefly discuss, what are the current challenges faced by data scientists and ML experts while working on these models.
- Unavailability of the proper quality dataset: As we know the Machine Learning model gets insights from data, having poor quality data for training impacts the accuracy of the model.
- Overfitting and Underfitting: A machine learning model should be able to generalize the results rather than cramming the dataset.
Overfitting: It occurs when a machine learning model has good accuracy on training data but performs very badly on the test dataset.
Underfitting: It occurs when the model is unable to get good accuracy on both training and test dataset.
3. Cost and time constraints: Training a machine learning model can be quite expensive and resource-intensive because we need high computational hardware to train the data.
Tryout:
Just to get familiar with the capabilities of Machine Learning, you can go through some cool APIs developed by Microsoft, one of them is
Computer Vision API :
Just go to this link and check out how well the Machine Learning model helps in recognizing the various objects in the image.
I hope you liked this article, please do give claps and follow us, if the blog has added some value to your learnings so far. Feel Free to reach out to me at medha.rwt@gmail.com for any queries related to content. Suggestions for any betterment of this article are highly welcome. Do let me know your reviews about this blog. And tell me the next topic you want me to write for you guys.
Till then Stay Safe & Keep Reading.
